When Safari users in Hong Kong recently tried to load the popular code-sharing website GitLab, they received a strange warning instead: Apple’s browser was blocking the site for their own safety. The access was temporarily cut off thanks to Apple’s use of a Chinese corporate website blacklist, which resulted in the innocuous site being flagged as a purveyor of misinformation. Neither Tencent, the massive Chinese firm behind the web filter, nor Apple will say how or why the site was censored.

The outage was publicized just ahead of the new year. On December 30, 2022, Hong Kong-based software engineer and former Apple employee Chu Ka-cheong tweeted that his web browser had blocked access to GitLab, a popular repository for open-source code. Safari’s “safe browsing” feature greeted him with a full-page “deceptive website warning,” advising that because GitLab contained dangerous “unverified information,” it was inaccessible. Access to GitLab was restored several days later, after the situation was brought to the company’s attention.

The warning screen itself came courtesy of Tencent, the mammoth Chinese internet conglomerate behind WeChat and League of Legends. The company operates the safe browsing filter for Safari users in China on Apple’s behalf — and now, as the Chinese government increasingly asserts control of the territory, in Hong Kong as well.

Apple spokesperson Nadine Haija would not answer questions about the GitLab incident, suggesting they be directed at Tencent, which also declined to offer responses.

The episode raises thorny questions about privatized censorship done in the name of “safety” — questions that neither company seems interested in answering: How does Tencent decide what’s blocked? Does Apple have any role? Does Apple condone Tencent’s blacklist practices?

“They should be responsible to their customers in Hong Kong and need to describe how they will respond to demands from the Chinese authorities to limit access to information,” wrote Charlie Smith, the pseudonymous founder of GreatFire, a Chinese web censorship advocacy and watchdog group. “Presumably people purchase Apple devices because they believe the company when they say that ‘privacy is a fundamental human right’. What they fail to add is *except if you are Chinese.”

Ka-cheong tweeted that other Hong Kong residents had reported GitLab similarly blocked on their devices thanks to Tencent. “We will look into it,” Apple engineer Maciej Stachowiak tweeted in response. “Thanks for the heads-up.” But Ka-cheong, who also serves as vice president of Internet Society Hong Kong Chapter, an online rights group, said he received no further information from Apple.

“Presumably people purchase Apple devices because they believe the company when they say that ‘privacy is a fundamental human right’. What they fail to add is *except if you are Chinese.”

The block came as a particular surprise to Ka-cheong and other Hong Kong residents because Apple originally said the Tencent blocklist would be used only for Safari users inside mainland China. According to a review of the Internet Archive, however, sometime after November 24, 2022, Apple quietly edited its Safari privacy policy to note that the Tencent blacklist would be used for devices in Hong Kong as well. (Haija, the Apple spokesperson, did not respond when asked when or why Apple expanded the use of Tencent’s filter to Hong Kong.)

Though mainland China has heavily censored internet access for decades, Hong Kong typically enjoyed unfettered access to the web, a freedom only recently threatened by the passage of a sweeping, repressive national security law in 2020.

Silently expanding the scope of the Tencent list not only allows Apple to remain in the good graces of China — whose industrial capacity remains existentially vital to the California-based company — but also provides plausible deniability about how or why such site blocks happen.

“While unfortunately many tech companies proactively apply political and religious censorship to their mainland Chinese users, Apple may be unique among North American tech companies in proactively applying such speech restrictions to users in Hong Kong,” said Jeffrey Knockel, a researcher with Citizen Lab, a digital security watchdog group at the University of Toronto.

Knockel pointed out that while a company like Tencent should expected to comply with Chinese law as a matter of course, Apple has gone out of its way to do so.

“The aspect which we should be surprised by and concerned about is Apple’s decision to work with Tencent in the first place to filter URLs for Apple’s Hong Kong users,” he said, “when other North American tech companies have resisted Hong Kong’s demands to subject Hong Kong users to China-based filtering.”

The block on GitLab would not be the first time Tencent deemed a foreign website “dangerous” for apparently ideological reasons. In 2020, attempts to visit the official website of Notepad++, a text editor app whose French developer had previously issued a statement of solidarity with Hong Kong dissidents, were blocked for users of Tencent web browsers, again citing safety.

The GitLab block also wouldn’t be the first time Apple, which purports to hold deep commitments to human rights, has bent the company’s products to align with Chinese national pressure. In 2019, Apple was caught delisting an app Hong Kong political dissidents were using to organize; in November, users noticed the company had pushed a software update to Chinese iPhone users that significantly weakened the AirDrop feature, which protesters throughout the country had been using to spread messages on the ground.

“All companies have a responsibility to respect human rights, including freedom of expression, no matter where in the world they operate,” Michael Kleinman, head of Amnesty International’s Silicon Valley Initiative, wrote to The Intercept. “Any steps by Apple to limit freedom of expression for internet users in Hong Kong would contravene Apple’s responsibility to respect human rights under the UN Guiding Principles.”

In 2019, Apple publicly acknowledged that it had begun using a “safe browsing” database maintained by Tencent to filter the web activity of its users in China, instead of an equivalent list operated by Google. Safe browsing filters ostensibly protect users from malicious pages containing malware or spear-phishing attacks by checking the website they’re trying to load against a master list of blacklisted domains.

In order to make such a list work, however, at least some personal information needs to be transmitted to the company operating the filter, be it Google or Tencent. When news of Apple’s use of the Tencent safe browsing list first broke, Matthew Green, a professor of cryptography at Johns Hopkins University, described it as “another example of Apple making significant modifications to its privacy infrastructure, largely without publicity or announcement.”

“I suppose the nature of having a ‘misinformation’ category is that China is going to have its own views on what that means.”

While important questions remain about exactly what information from Safari users in Hong Kong and China is ultimately transmitted to Tencent and beyond, the GitLab incident shows another troubling aspect of safe browsing: It gives a single company the ability to unilaterally censor the web under the aegis of public safety.

“Our concern was that outsourcing this stuff to Chinese firms seemed problematic for Apple,” Green explained in an interview with The Intercept, “and I suppose the nature of having a ‘misinformation’ category is that China is going to have its own views on what that means.”

Indeed, it’s impossible to know in what sense GitLab could have possibly been considered a source of dangerous “unverified information.” The site is essentially an empty vessel where software developers, including corporate clients like T-Mobile and Goldman Sachs, can safely store and edit code. The Chinese government has recently cracked down on some open-source code sites similar to GitLab, where engineers from around the world are able to freely interact, collaborate, and share information. (GitLab did not respond to a request for comment.)

Notably, the censorship-evasion and anonymity web browser Tor has turned to GitLab to catalog instances of Chinese state internet censorship, though there’s no indication it was this activity that led to GitLab’s addition to the Tencent list.

While Tencent provides some public explanation of its criteria for blocking a website, its decision-making process is completely opaque, and the published censorship standards are extremely vague, including offenses like “endangering national security” and “undermining national unity.”

Tencent has long been scrutinized for its ties to the Chinese government, which frequently leverages state power to more closely influence or outright control nominally private firms.

Earlier this month, the Financial Times reported that the Chinese government was acquiring so-called golden shares of Tencent, a privileged form of equity that’s become “a common tool used by the state to exert influence over private news and content companies.” A 2021 New York Times report on Tencent noted the company’s eagerness to cooperate with Chinese government mandates, quoting the company’s president during an earnings call that year: “Now I think it’s important for us to understand even more about what the government is concerned about, what the society is concerned about, and be even more compliant.”

While Tencent’s compliance with the Chinese national security agenda ought not to come as a surprise, Knockel of Citizen Lab says Apple’s should.

“Ultimately I don’t think it really matters exactly how GitLab came to be blocked by Tencent’s Safe Browsing,” he said. “Tencent’s blocking of GitLab for Safari users underscores that Apple’s subjection of Hong Kong users to screening via a China-based company is problematic not only in principle but also in practice.”

The post Apple Brings Mainland Chinese Web Censorship to Hong Kong appeared first on The Intercept.

This post was originally published on The Intercept.

Read this story in Persian

As furious anti-government protests swept Iran, the authorities retaliated with both brute force and digital repression. Iranian mobile and internet users reported rolling network blackouts, mobile app restrictions, and other disruptions. Many expressed fears that the government can track their activities through their indispensable and ubiquitous smartphones.

Iran’s tight grip on the country’s connection to the global internet has proven an effective tool for suppressing unrest. The lack of clarity about what technological powers are held by the Iranian government — one of the most opaque and isolated in the world — has engendered its own form of quiet terror for prospective dissidents. Protesters have often been left wondering how the government was able to track down their locations or gain access to their private communications — tactics that are frighteningly pervasive but whose mechanisms are virtually unknown.

“This is not a surveillance system but rather a repression and control system to limit the capability of users to dissent or protest.”

While disconnecting broad swaths of the population from the web remains a favored blunt instrument of Iranian state censorship, the government has far more precise, sophisticated tools available as well. Part of Iran’s data clampdown may be explained through the use of a system called “SIAM,” a web program for remotely manipulating cellular connections made available to the Iranian Communications Regulatory Authority. The existence of SIAM and details of how the system works, reported here for the first time, are laid out in a series of internal documents from an Iranian cellular carrier that were obtained by The Intercept.

According to these internal documents, SIAM is a computer system that works behind the scenes of Iranian cellular networks, providing its operators a broad menu of remote commands to alter, disrupt, and monitor how customers use their phones. The tools can slow their data connections to a crawl, break the encryption of phone calls, track the movements of individuals or large groups, and produce detailed metadata summaries of who spoke to whom, when, and where. Such a system could help the government invisibly quash the ongoing protests — or those of tomorrow — an expert who reviewed the SIAM documents told The Intercept.

“SIAM can control if, where, when, and how users can communicate,” explained Gary Miller, a mobile security researcher and fellow at the University of Toronto’s Citizen Lab. “In this respect, this is not a surveillance system but rather a repression and control system to limit the capability of users to dissent or protest.”

SIAM gives the government’s Communications Regulatory Authority — Iran’s telecommunications regulator — turnkey access to the activities and capabilities of the country’s mobile users. “Based on CRA rules and regulations all telecom operators must provide CRA direct access to their system for query customers information and change their services via web service,” reads an English-language document obtained by The Intercept. (Neither the CRA nor Iran’s mission to the United Nations responded to a requests for comment.)

The SIAM documents are drawn from a trove of internal materials from the Iranian cellular carrier Ariantel, including years of email correspondence and a variety of documents shared between Ariantel employees, outside contractors, and Iranian government personnel. The cache of materials was shared with The Intercept by an individual who claimed to have hacked Ariantel, and believed the documents were in the public interest given the ongoing protests in Iran and the threat SIAM might pose to demonstrators. (Ariantel did not respond to a request for comment.)

The details of the program reported here are drawn largely from two documents contained in the archive. The first is a Persian-language user manual for SIAM that appears to have originated from within the Office of Security of Communications Systems, or OSCS, a subdivision of the CRA. Emails reviewed by The Intercept show that this SIAM manual was sent to Ariantel directly by the CRA and repeatedly forwarded between the mobile carrier’s employees in recent years. The emails show that the CRA and Ariantel discussed SIAM as recently as August. The second document, produced during a proposed deal with a Spanish telecom contractor, is an English-language manual that documents many of the same SIAM capabilities. Miller told The Intercept that the English SIAM manual appeared to be written by a person or people with specialized technical knowledge of mobile networks.

Experts on mobile security and Iranian government censorship say the functionality revealed by the SIAM program poses a clear threat to protesters demonstrating against the government over the past month.

“These functions can lead to life-and-death situations in a country like Iran, where there is no fair judicial process, no accountability, and we have a huge pattern of violations of people’s rights,” said Amir Rashidi, an internet security and digital rights expert focused on Iran. “Using the tools outlined in this manual could not only lead to mass surveillance and violations of privacy — it can also easily be used to identify the location of protesters who are literally risking their lives to fight for their basic rights.”

Iranians regularly complain of slowed internet access on mobile devices during periods of protest — an abrupt dip in service that makes smartphone usage difficult if not impossible at moments when such a device could be crucial. Based on the manuals, SIAM offers an effortless way to throttle a phone’s data speeds, one of roughly 40 features included in the program. This ability to downgrade users’ speed and network quality is particularly pernicious because it can not only obstruct one’s ability to use their phone, but also make whatever communication is still possible vulnerable to interception.

Referred to within SIAM as “Force2GNumber,” the command allows a cellular carrier to kick a given phone off substantially faster, more secure 3G and 4G networks and onto an obsolete and extremely vulnerable 2G connection. Such a network downgrade would simultaneously render a modern smartphone largely useless and open its calls and texts to interception — both of obvious utility to a government clamping down on public gatherings and speech.

While not directly mentioned in the manuals, downgrading users to a 2G connection could also expose perilously sensitive two-factor authentication codes delivered to users through SMS. The Iranian government has previously attempted to undermine two-factor authentication, including through malware campaigns targeting dissidents.

“Generally speaking, forcing a phone to use the 2G network would still allow the phone to receive a two-factor SMS authentication message because SMS is sent over the mobile signaling network,” explained Miller. “However, the effect of forcing a user onto the 2G network, more importantly, would essentially render the corresponding real-time application services such as P2P communication, social media, and internet useless.”

While current 5G and 4G cellular connections have more robust built-in encryption systems to thwart eavesdropping, the 2G cellular standard, first introduced in 1991, generally does not encrypt data or uses outdated encryption methods that are easy to crack. Law enforcement agencies in the United States have also employed this technique, using hardware like the controversial “stingray” device to create a bogus 2G network blanketing a small area and then trick targeted phones into connecting to it.

Miller pointed out that the target of a 2G downgrade might experience the attack as little more than spotty cell reception. “It can be viewed as a method to appear as if the network is congested and severely limit a user’s data services,” Miller said.

Slowing connectivity is only one of many telecom tools available to Ariantel — and the CRA — that could be used to monitor political dissent. SIAM also provides a range of tools to track the physical locations of cell users, allowing authorities to both follow an individual’s movements and identify everyone present at a given spot. Using the “LocationCustomerList” command allows SIAM operators to see what phone numbers have connected to specified cell towers along with their corresponding IMEI number, a unique string of numbers assigned to every mobile phone in the world. “For example,” Miller said, “if there is a location where a protest is occurring, SIAM can provide all of the phone numbers currently at that location.”

SIAM’s tracking of unique device identifiers means that swapping SIM cards, a common privacy-preserving tactic, may be ineffective in Iran since IMEI numbers persist even with a new SIM, explained a network security researcher who reviewed the manuals and spoke on the condition of anonymity, citing their safety.

SIAM’s location-tracking power is particularly alarming given the high-stakes protests taking place across Iran. The Intercept reviewed undated text messages sent to Iranian mobile phone users from local police in the city of Isfahan informing them that they had been confirmed to have been in a location of “unrest” and warning them not to attend in the future. Many Iranian social media users have reported receiving similar messages in recent weeks, warning them to stay away from the scene of protests or from associating with “anti-revolutionary” opponents of the government online.

Armed with a list of offending phone numbers, SIAM would make it easy for the Iranian government to rapidly drill down to the individual level and pull a vast amount of personal information about a given mobile customer, including where they’ve been and with whom they’ve communicated. According to the manuals, user data accessible through SIAM includes the customer’s father’s name, birth certificate number, nationality, address, employer, billing information, and location history, including a record of Wi-Fi networks and IP addresses from which the user has connected to the internet.

While much of Iran’s surveillance capacity remains shrouded in mystery, details about the SIAM program contained in the Ariantel archive provide a critical window into the types of tools the Iranian government has at its disposal to monitor and control the internet, as it confronts what may be the greatest threat to its rule in decades.

“These documents prove something that we have long suspected, which is that even devices that use encryption for messaging are still vulnerable because of the nature of internet infrastructure in Iran,” said Mahsa Alimardani, a senior researcher with the internet freedom organization Article 19. “Security measures like two-factor identification using text messages still depend on telecommunications companies connected to the state. Average internet users are forced to connect through nodes controlled by these companies, and their centralization of authority with the government makes users vulnerable to insidious types of surveillance and control.”

The latest round of protests in Iran kicked off in mid-September, after a young woman named Mahsa Jina Amini was killed while in the custody of the country’s notorious morality police, following her arrest for wearing her mandatory head covering improperly. While the movement originated with women opposing the brutality of hijab enforcement, anti-government outrage quickly spread among Iran’s youth, from universities to secondary schools across the country. The government’s crackdown took a variety of shapes, including brute force, with security services in riot gear squaring off with demonstrators in the street and a quieter effort to shut down civilian communications.

Internet shutdowns have by now become a familiar tool of political control in the hands of the Iranian government and other states. A violent Iranian crackdown against protests over fuel prices in November 2019 was accompanied by a nationwide shutdown lasting nearly a week, the first-ever use of an internet blackout to isolate an entire country. That shutdown severed tens of millions of people from the global internet. It was a chilling demonstration of the broad technical powers that Iranian authorities had quietly engineered.

The CRA is known to play an integral role in filtering Iran’s internet access. In 2013, the agency was among a list of Iranian government entities sanctioned by the U.S. Treasury Department for its role in the “blockage of hundreds of public Internet websites” around the time of the disputed 2009 Iranian presidential election. The agency’s powers are believed to have grown since then, as the Iranian government has embraced the concept of “internet sovereignty” as a means of social control. A report on the November 2019 cyber crackdown by Article 19 found that the shutdowns were carried out in large part by officials from the CRA ordering internet service providers to shut down during the unrest.

The Iranian government has long viewed internet freedom as a national security issue and has taken steps to securitize Iranians’ online access. As in the United States, where the National Security Agency has used government secrecy and legal coercion to turn the telecom and data sectors into intelligence-gathering tools, the Iranian state compels communications networks to give the government access through required hardware and software. In Iran, where the autocratic reach of central government leadership touches nearly every aspect of the state without even superficial democratic oversight, the powers afforded by this integration are far greater and far more draconian in consequence.

Part of this effort has included directly assigning Iranian intelligence personnel to government bodies tasked with internet regulation, like the CRA. The Article 19 report notes the close personnel relationship between the CRA’s OSCS division and Iran’s Ministry of Intelligence.

Though Iranians have complained of slowed data connections and total internet blackouts at times, the telecom crackdown has consequences beyond losing one’s connection. Demonstrators have reported visits from government authorities at their homes, where the agents were armed with specific knowledge of their whereabouts and activities, such as when they were using their phones to record video.

While some of what SIAM does is benign and required for administrating any cellular network, Miller, the Citizen Lab researcher, explained that the scope of the system and the Iranian government’s access to it is not. While most countries allow law enforcement and security agencies to legally obtain, intercept, and analyze cellular communications, the surveillance and control powers afforded by SIAM are notable in their scale and degree, said Miller: “The requests by CRA go well beyond traditional lawful intercept requirements, at least in non-repressive countries.”

SIAM allows its operators to learn a great deal not just about where a customer has been, but also what they’ve been up to, a bounty of personal data that, Miller said, “can enable CRA to create a social network/profile of the user based on his/her communication with other people.”

“Controlling user communications is a massive violation of basic and fundamental human rights.”

By entering a particular phone number and the command “GetCDR” into SIAM, a system user can generate a comprehensive Call Detail Record, including the date, time, duration, location, and recipients of a customer’s phone calls during a given time period. A similar rundown can be conducted for internet usage as well using the “GetIPDR” command, which prompts SIAM to list the websites and other IP addresses a customer has connected to, the time and date these connections took place, the customer’s location, and potentially the apps they opened. Such a detailed record of internet usage could also reveal users running virtual private networks, which are used to cover a person’s internet trail by routing their traffic through an encrypted connection to an outside server. VPNs — including some banned by the government — have become tremendously popular in Iran as a means of evading domestic web censorship.

Though significantly less subtle than being forced onto a 2G network, SIAM can also be used to entirely pull the plug on a customer’s device at will. Through the “ApplySuspIp” command, the system can entirely disconnect any mobile phone on the network from the internet for predetermined lengths of time or permanently. Similar commands would let SIAM block a user from placing or receiving calls.

Rashidi, the internet security expert, said participants in the recent demonstrations, as well as Iranians living near scenes of protest, have reported internet shutdowns targeting their mobile devices that have downgraded phones to 2G access, particularly during the late afternoons and evenings when many demonstrations occur.

Rashidi said the widespread use of VPNs in Iran represents another vulnerability the SIAM system could exploit. The program makes it possible to check particular IP addresses against particular VPNs and thereby deduce the identities and locations of the users accessing them. “The government can easily identify IP addresses in use by a particular VPN provider, pass the addresses to this location function, and then see where the people are who are using this VPN,” said Rashidi.

Although the documents don’t mention SIAM’s use against protesters or any other specific target, Miller said the functionality matches what he’s observed in this and other digital crackdowns in Iran. “CRA has defined rules and regulations to provide direct access to mobile operators’ system, and SIAM is a means to this end,” he said. “If all telecom operators in Iran are required to provide the CRA with SIAM or similar direct access, they could, in effect have complete control over all user mobile communications throughout the country. Controlling user communications is a massive violation of basic and fundamental human rights.”

The post Iran’s Secret Manual for Tracking and Controlling Protesters’ Mobile Phones appeared first on The Intercept.

This post was originally published on The Intercept.

Elon Musk is once again suggesting his business interests can solve a high-profile crisis: This time, the SpaceX CEO says Starlink satellite internet can alleviate Iran’s digital crackdown against ongoing anti-government protests. Iranian dissidents and their supporters around the world cheered Musk’s announcement that Starlink is now theoretically available in Iran, but experts say the plan is far from a censorship panacea.

Musk’s latest headline-riding gambit came after Iran responded to the recent rash of nationwide protests with large-scale disruption of the country’s internet access. On September 23, Secretary of State Anthony Blinken announced the U.S. was easing restrictions on technology exports to help counter Iranian state censorship efforts.

Musk, ready to pounce, quickly replied: “Activating Starlink …”

Predictably, Musk’s dramatic tweet set off a frenzy. Within a day, venture capitalist and longtime Musk-booster Shervin Pishevar was already suggesting Musk had earned the Nobel Peace Prize. Just the thought of Starlink “activating” an uncensored internet for millions during a period of Middle Eastern political turmoil was an instant public relations coup for Musk.

In Iran, though, the notion of a benevolent American billionaire beaming freedom to Iran by satellite is derailed by the demands of reality, specifically physics. Anyone who wants to use Starlink, the satellite internet service provider operated by Musk’s rocketry concern, SpaceX, needs a special dish to send and receive internet data.

“I don’t think it’s much of a practical solution because of the problem of smuggling in the ground terminals.”

While it may be possible to smuggle Starlink hardware into Iran, getting a meaningful quantity of satellite dishes into Iran would be an incredible undertaking, especially now that the Iranian government has been tipped off to the plan on Twitter.

Todd Humphreys, an engineering professor at the University of Texas at Austin whose research focuses on satellite communication, said, “I don’t think it’s much of a practical solution because of the problem of smuggling in the ground terminals.”

The idea is not without precedent. In Ukraine, after the Russian invasion disrupted internet access, the deployment of Musk’s satellite dishes earned him international press adulation and a bevy of lucrative government contracts. In Ukraine, though, Starlink was welcomed by a profoundly pro-American government desperate for technological aid from the West. U.S. government agencies were able to ship the requisite hardware with the full logistical cooperation of the Ukrainian government.

This is not, to say the very least, the case in Iran, where the government is unlikely to condone the import of a technology explicitly meant to undermine its own power. While Musk’s claim that Starlink’s orbiting satellites are activated over Iran may be true, the notion that censorship-free internet connectivity is something that can be flipped on like a light switch is certainly not. Without dishes on the ground to communicate with the satellites, it’s a meaningless step: technologically tantamount to giving a speech to an empty room.

Humphreys, who has previously done consulting work for Starlink, explained that because of the specialized nature of Starlink hardware, it’s doubtful Iranians could craft a DIY alterative. “It’s not like you can build a homebrew receiver,” he said. “It’s a very complicated signal structure with a very wideband signal. Even a research organization would have a hard time.”

Musk is famously uninterested in the constraints imposed by reality, but he seems to acknowledge the problem to some degree. In a September 25 tweet, Carnegie Endowment for International Peace fellow Karim Sadjadpour wrote, “I spoke w/ @elonmusk about Starlink in Iran, he gave me permission to share this: ‘Starlink is now activated in Iran. It requires the use of terminals in-country, which I suspect the [Iranian] government will not support, but if anyone can get terminals into Iran, they will work.’”

Implausibility hasn’t stopped Musk’s fans, either. One tweet from a senior fellow at the Atlantic Council purporting to document a Starlink dish already successfully secreted into Iran turned out to be a photo from 2020, belonging to an Idaho man who happened to have a Persian rug.

The fandom — and the starpower it’s attached to — might be the point here. Given the obstacles, Musk’s Starlink aspirations may be best understood in the context of his past spectacular, spectacularly unfulfilled claims, rather than something akin to Starlink’s rapid adoption in Ukraine. Musk’s penchant for internet virality has become a key component of his business operations. He has repeatedly made bold pronouncements, typically on Twitter, that a technology he happens to manufacture is the key to cracking some global crisis. Whether it’s Thai children stuck in a waterlogged cave, the Covid-19 pandemic, or faltering American transit infrastructure, Musk has repeatedly offered technological solutions that are either plainly implausible, botched in execution, or a mixture of both.

It’s not just the lack of dishes in Iranian homes. Musk’s plan is further complicated by Starlink’s reliance on ground stations: communications facilities that allow the SpaceX satellites to plug into earthbound internet infrastructure from orbit. While upgraded Starlink satellites may no longer need these ground stations in the near future, the network of today still largely requires them to service a country as vast as Iran, said Humphreys, the University of Texas professor. Again, Iran is unlikely to approve the construction within its borders of satellite installations owned by an American defense contractor.

Humphreys suggested that ground stations built in a neighboring country could provide some level of connection, albeit at reduced speed, but that still doesn’t get over the hump of every Iranian who wants to get online needing a $550 kit with “Starlink” emblazoned on the box. While Humphreys added that he was hopeful that a slow trickle of Starlinks terminals could aid Iranian dissidents over time, he said, “I don’t think in the short term this will have an impact on the unrest in Iran.”

Alp Toker, director of the internet monitoring and censorship watchdog group NetBlocks, noted that many Iranians already watch banned satellite television channels through contraband dishes, meaning the smuggling of Starlink dishes is doable in theory. While he praised the idea of bringing Starlink to Iran as “credible and worthwhile” in the long term, the difficulty in sourcing Starlink’s specialized equipment means that accessing Musk’s satellites remains “a solution for the few,” not a counter to population-scale censorship.

While future versions of the Starlink system might be able to communicate with more accessible devices like handheld phones, Toker said, “As far as we know this isn’t possible with the current generation of kit, and it won’t be until then that Starlink or similar platforms could simply ‘switch on’ internet in a country in the sense that most people understand.”

Even with Iran’s culture of bootleg satellite TV, these experts warned that a Starlink connection could endanger Iranians. Rose Croshier, a policy fellow at the Center for Global Development, noted the risks: “A word of caution: TV dishes are passive — they don’t transmit — so a Starlink terminal (that both receives and transmits data) in a crowd of illegal satellite dishes would still be very findable by Iranian authorities.”

“I don’t think in the short term this will have an impact on the unrest in Iran.”

The plan faces further terrestrial hurdles. The complex two-way nature of satellite connections is part of why they’re subject to international regulation, most notably through the International Telecommunication Union, of which both the United States and Iran are members. Croshier pointed to a 2021 paper on satellite internet usage by the Asia Development Bank that explained how “US-based entities such as Starlink … require regulatory approval from the FCC as well the ITU” and that “service provision to customers will require regulatory approval in every country of operation.” Mahsa Alimardani, a senior Middle East researcher at Article19, a free expression advocacy group, tweeted that even if Starlink could beam internet to Iranians in a meaningful way, the company would face consequences from the International Telecommunications Union if it did so without Iranian approval — approval it is unlikely to ever get.

Then there are sanctions against Iran. Blinken, the secretary of state, announced a relaxation of tech exports, but the restrictions on trade with Iran remain a serious obstacle. “There are a host of human rights related sanctions on Iranian actors in the IT space under a sanctions authority called GHRAVITY that complicate any of this beyond the questions raised of whether Iran would allow Starlink terminals in country,” explained Brian O’Toole, a senior fellow at the Atlantic Council and expert on global sanctions. The relaxed rules would still require a special license for Starlink use in Iran, O’Toole said, which he doubts would be granted: “Much of this Starlink stuff doesn’t appear terribly likely to do much, from my point of view.”

Starlink — or a competitor — may one day bring unfettered net uplinks to Iran and other countries where online dissent is choked out, but for today’s Iranian protesters, the realities far exceed the PR punch of a two-word tweet.

The post No, Elon Musk’s Starlink Probably Won’t Fix Iranian Internet Censorship appeared first on The Intercept.

This post was originally published on The Intercept.

Facebook and Instagram’s speech policies harmed fundamental human rights of Palestinian users during a conflagration that saw heavy Israeli attacks on the Gaza Strip last May, according to a study commissioned by the social media sites’ parent company Meta.

“Meta’s actions in May 2021 appear to have had an adverse human rights impact … on the rights of Palestinian users to freedom of expression, freedom of assembly, political participation, and non-discrimination, and therefore on the ability of Palestinians to share information and insights about their experiences as they occurred,” says the long-awaited report, which was obtained by The Intercept in advance of its publication.

Commissioned by Meta last year and conducted by the independent consultancy Business for Social Responsibility, or BSR, the report focuses on the company’s censorship practices and allegations of bias during bouts of violence against Palestinian people by Israeli forces last spring.

“Meta’s actions in May 2021 appear to have had an adverse human rights impact.”

Following protests over the forcible eviction of Palestinian families from the Sheikh Jarrah neighborhood in occupied East Jerusalem, Israeli police cracked down on protesters in Israel and the West Bank, and launched military air strikes against Gaza that injured thousands of Palestinians, killing 256, including 66 children, according to the United Nations. Many Palestinians attempting to document and protest the violence using Facebook and Instagram found their posts spontaneously disappeared without recourse, a phenomenon the BSR inquiry attempts to explain.

Last month, over a dozen civil society and human rights groups wrote an open letter protesting Meta’s delay in releasing the report, which the company had originally pledged to release in the “first quarter” of the year.

While BSR credits Meta for taking steps to improve its policies, it further blames “a lack of oversight at Meta that allowed content policy errors with significant consequences to occur.”

Though BSR is clear in stating that Meta harms Palestinian rights with the censorship apparatus it alone has constructed, the report absolves Meta of “intentional bias.” Rather, BSR points to what it calls “unintentional bias,” instances “where Meta policy and practice, combined with broader external dynamics, does lead to different human rights impacts on Palestinian and Arabic speaking users” — a nod to the fact that these systemic flaws are by no means limited to the events of May 2021.

Meta responded to the BSR report in a document to be circulated along with the findings. (Meta did not respond to The Intercept’s request for comment about the report by publication time.) In a footnote in the response, which was also obtained by The Intercept, the company wrote, “Meta’s publication of this response should not be construed as an admission, agreement with, or acceptance of any of the findings, conclusions, opinions or viewpoints identified by BSR, nor should the implementation of any suggested reforms be taken as admission of wrongdoing.”

According to the findings of BSR’s report, Meta deleted Arabic content relating to the violence at a far greater rate than Hebrew-language posts, confirming long-running complaints of disparate speech enforcement in the Palestinian-Israeli conflict. The disparity, the report found, was perpetuated among posts reviewed both by human employees and automated software.

“The data reviewed indicated that Arabic content had greater over-enforcement (e.g., erroneously removing Palestinian voice) on a per user basis,” the report says. “Data reviewed by BSR also showed that proactive detection rates of potentially violating Arabic content were significantly higher than proactive detection rates of potentially violating Hebrew content.”

BSR attributed the vastly differing treatment of Palestinian and Israeli posts to the same systemic problems rights groups, whistleblowers, and researchers have all blamed for the company’s past humanitarian failures: a dismal lack of expertise. Meta, a company with over $24 billion in cash reserves, lacks staff who understand other cultures, languages, and histories, and is using faulty algorithmic technology to govern speech around the world, the BSR report concluded.

Not only do Palestinian users face an algorithmic screening that Israeli users do not — an “Arabic hostile speech classifier” that uses machine learning to flag potential policy violations and has no Hebrew equivalent — the report notes that the Arabic system also doesn’t work well: “Arabic classifiers are likely less accurate for Palestinian Arabic than other dialects, both because the dialect is less common, and because the training data — which is based on the assessments of human reviewers — likely reproduces the errors of human reviewers due to lack of linguistic and cultural competence.”

Human employees appear to have exacerbated the lopsided effects of Meta’s speech-policing algorithms. “Potentially violating Arabic content may not have been routed to content reviewers who speak or understand the specific dialect of the content,” the report says. It also notes that Meta didn’t have enough Arabic and Hebrew-speaking staff on hand to manage the spike in posts.

These faults had cascading speech-stifling effects, the report continues. “Based on BSR’s review of tickets and input from internal stakeholders, a key over-enforcement issue in May 2021 occurred when users accumulated ‘false’ strikes that impacted visibility and engagement after posts were erroneously removed for violating content policies.” In other words, wrongful censorship begat further wrongful censorship, leaving the affected wondering why no one could see their posts. “The human rights impacts … of these errors were more severe given a context where rights such as freedom of expression, freedom of association, and safety were of heightened significance, especially for activists and journalists,” the report says.

Beyond Meta’s failures in triaging posts about Sheikh Jarrah, BSR also points to the company’s “Dangerous Individuals and Organizations” policy — referred to as “DOI” in the report — a roster of thousands of people and groups that Meta’s billions of users cannot “praise,” “support,” or “represent.” The full list, obtained and published by The Intercept last year, showed that the policy focuses mostly on Muslim and Middle Eastern entities, which critics described as a recipe for glaring ethnic and religious bias.

Meta claims that it’s legally compelled to censor mention of groups designated by the U.S. government, but legal scholars have disputed the company’s interpretation of federal anti-terrorism laws. Following The Intercept’s report on the list, the Brennan Center for Justice called the company’s claims of legal obligation a “fiction.”

“Meta’s DOI policy and the list are more likely to impact Palestinian and Arabic-speaking users, both based upon Meta’s interpretation of legal obligations, and in error.”

BSR agrees the policy is systemically biased: “Legal designations of terrorist organizations around the world have a disproportionate focus on individuals and organizations that have identified as Muslim, and thus Meta’s DOI policy and the list are more likely to impact Palestinian and Arabic-speaking users, both based upon Meta’s interpretation of legal obligations, and in error.”

Palestinians are particularly vulnerable to the effects of the blacklist, according to the report: “Palestinians are more likely to violate Meta’s DOI policy because of the presence of Hamas as a governing entity in Gaza and political candidates affiliated with designated organizations. DOI violations also come with particularly steep penalties, which means Palestinians are more likely to face steeper consequences for both correct and incorrect enforcement of policy.”

The document concludes with a list of 21 nonbinding policy recommendations, including increased staffing capacity to properly understand and process Arabic posts, implementing a Hebrew-compatible algorithm, increased company oversight of outsourced moderators, and both reforms to and increased transparency around the “Dangerous Individuals and Organizations” policy.

In its response to the report, Meta vaguely commits to implement or consider implementing aspects of 20 out of 21 the recommendations. The exception is a call to “Fund public research into the optimal relationship between legally required counterterrorism obligations and the policies and practices of social media platforms,” which the company says it will not pursue because it does not wish to provide legal guidance for other companies. Rather, Meta suggests concerned experts reach out directly to the federal government.

The post Facebook Report Concludes Company Censorship Violated Palestinian Human Rights appeared first on The Intercept.

This post was originally published on The Intercept.

A Twitter spokesperson said the social media giant deleted a Carnegie Mellon University professor’s controversial tweet condemning Queen Elizabeth II on the grounds that it was “abusive.” The company defines abusive behavior as “an attempt to harass, intimidate, or silence someone else’s voice” — in this case the voice of the world’s longest-reigning monarch.

It was a banner day for posting. As soon as news of the queen’s impending death hit Twitter, the platform was quickly dominated by a global outpouring of both grief and glee, a heated mixture of paeans to the queen’s 70-year tenure and angry denunciations of the British monarchy’s legacy of colonial violence and exploitation. Among the latter was Carnegie Mellon’s Uju Anya, an associate professor of second language acquisition. “I heard the chief monarch of a thieving raping genocidal empire is finally dying. May her pain be excruciating,” Anya tweeted.

“We took enforcement action on the account you referenced for violating the Twitter Rules on abusive behaviour,” Twitter spokesperson Lauren Myers-Cavanagh, using the British spelling of “behavior,” wrote to The Intercept in response to a query.

“It does highlight the power imbalances that can often exist in the way these platforms treat powerful figures.”

The removal of the post illustrates how criticisms of powerful people, however distasteful, can be disappeared from social media sites for murky reasons. “It does highlight the power imbalances that can often exist in the way these platforms treat powerful figures,” Evelyn Douek, an assistant professor at Stanford Law School and scholar of content moderation policies, told The Intercept. “Often people in power get allowances because it’s in the public interest but people don’t for criticizing them, even though that’s often clearly in the public interest too.”

Anya’s tweet immediately attracted widespread attention and criticism, not least because it was reproachfully quote-tweeted by Amazon founder Jeff Bezos. “This is someone supposedly working to make the world better?” the second-richest American tweeted. “I don’t think so. Wow.” Twitter users were quick to point out that Anya had recently tweeted approvingly of Chris Smalls, a rising labor leader instrumental in efforts to unionize Amazon warehouses.

After the criticisms came pouring in, Anya, who was born in Nigeria, tweeted in defense of her remarks: “If anyone expects me to express anything but disdain for the monarch who supervised a government that sponsored the genocide that massacred and displaced half my family and the consequences of which those alive today are still trying to overcome, you can keep wishing upon a star.” (Anya did not immediately respond to The Intercept’s request for comment.)

While the tweet was no doubt offensive to many fans of the crown in wishing suffering upon Elizabeth, the specific rule cited by Myers-Cavanagh justifying the censorship isn’t typically deployed in defense of royalty. The company claims that its ban on abusive remarks is designed to protect speech rather than delete it.

“In order to facilitate healthy dialogue on the platform, and empower individuals to express diverse opinions and beliefs, we prohibit behavior that harasses or intimidates, or is otherwise intended to shame or degrade others,” Twitter’s online help center says. “We consider abusive behavior an attempt to harass, intimidate, or silence someone else’s voice.”

Such speech is deleted to shield vulnerable voices from suppression. “On Twitter, you should feel safe expressing your unique point of view,” the policy reads. “We believe in freedom of expression and open dialogue, but that means little as an underlying philosophy if voices are silenced because people are afraid to speak up.”

It is not clear how the queen of England could ever be meaningfully “silenced” or “afraid to speak up” because of an academic’s tweet. Twitter did not respond in time for publication when asked whether the “abusive behavior” policy applies to the deceased.

The policy lays out a variety of “abusive” speech types, including “violent threats”; “content that wishes, hopes, promotes, incites, or expresses a desire for death, serious bodily harm or serious disease”; and “unwanted sexual advances.” It’s unclear which of these categories Anya’s tweet, published when its subject was on the verge of death and most likely not checking Twitter, would fall under.

The examples of abusive behavior that Twitter provides are of the more clear-cut “I hope you get cancer and die” variety, rather than an edge case hoping that someone already near death will experience greater suffering. The policy does note that in some cases offending tweets will be deleted while sparing the account from suspension, as was the case with Anya, when “regarding certain individuals credibly accused of severe violence,” an enforcement that would seem to implicitly agree with the professor’s underlying argument against the crown.

Douek, the Stanford professor, said that it seemed like an odd enforcement of the rule given the vast gulf in power between a professor and a monarch. “Unclear to me how the queen is going to be intimidated by that tweet,” she told The Intercept. “Surprised they stood by it, actually.”

The post Twitter Censored Professor’s Post for “Abusive Behaviour” Toward the Queen appeared first on The Intercept.

This post was originally published on The Intercept.

Training materials reviewed by The Intercept confirm that Google is offering advanced artificial intelligence and machine-learning capabilities to the Israeli government through its controversial “Project Nimbus” contract. The Israeli Finance Ministry announced the contract in April 2021 for a $1.2 billion cloud computing system jointly built by Google and Amazon. “The project is intended to provide the government, the defense establishment and others with an all-encompassing cloud solution,” the ministry said in its announcement.

Google engineers have spent the time since worrying whether their efforts would inadvertently bolster the ongoing Israeli military occupation of Palestine. In 2021, both Human Rights Watch and Amnesty International formally accused Israel of committing crimes against humanity by maintaining an apartheid system against Palestinians. While the Israeli military and security services already rely on a sophisticated system of computerized surveillance, the sophistication of Google’s data analysis offerings could worsen the increasingly data-driven military occupation.

According to a trove of training documents and videos obtained by The Intercept through a publicly accessible educational portal intended for Nimbus users, Google is providing the Israeli government with the full suite of machine-learning and AI tools available through Google Cloud Platform. While they provide no specifics as to how Nimbus will be used, the documents indicate that the new cloud would give Israel capabilities for facial detection, automated image categorization, object tracking, and even sentiment analysis that claims to assess the emotional content of pictures, speech, and writing. The Nimbus materials referenced agency-specific trainings available to government personnel through the online learning service Coursera, citing the Ministry of Defense as an example.

Jack Poulson, director of the watchdog group Tech Inquiry, shared the portal’s address with The Intercept after finding it cited in Israeli contracting documents.

“The former head of Security for Google Enterprise — who now heads Oracle’s Israel branch — has publicly argued that one of the goals of Nimbus is preventing the German government from requesting data relating on the Israel Defence Forces for the International Criminal Court,” said Poulson, who resigned in protest from his job as a research scientist at Google in 2018, in a message. “Given Human Rights Watch’s conclusion that the Israeli government is committing ‘crimes against humanity of apartheid and persecution’ against Palestinians, it is critical that Google and Amazon’s AI surveillance support to the IDF be documented to the fullest.”

Though some of the documents bear a hybridized symbol of the Google logo and Israeli flag, for the most part they are not unique to Nimbus. Rather, the documents appear to be standard educational materials distributed to Google Cloud customers and presented in prior training contexts elsewhere.

Google did not respond to a request for comment.

The documents obtained by The Intercept detail for the first time the Google Cloud features provided through the Nimbus contract. With virtually nothing publicly disclosed about Nimbus beyond its existence, the system’s specific functionality had remained a mystery even to most of those working at the company that built it. In 2020, citing the same AI tools, U.S Customs and Border Protection tapped Google Cloud to process imagery from its network of border surveillance towers.

Many of the capabilities outlined in the documents obtained by The Intercept could easily augment Israel’s ability to surveil people and process vast stores of data — already prominent features of the Israeli occupation.

“Data collection over the entire Palestinian population was and is an integral part of the occupation,” Ori Givati of Breaking the Silence, an anti-occupation advocacy group of Israeli military veterans, told The Intercept in an email. “Generally, the different technological developments we are seeing in the Occupied Territories all direct to one central element which is more control.”

The Israeli security state has for decades benefited from the country’s thriving research and development sector, and its interest in using AI to police and control Palestinians isn’t hypothetical. In 2021, the Washington Post reported on the existence of Blue Wolf, a secret military program aimed at monitoring Palestinians through a network of facial recognition-enabled smartphones and cameras.

“Living under a surveillance state for years taught us that all the collected information in the Israeli/Palestinian context could be securitized and militarized,” said Mona Shtaya, a Palestinian digital rights advocate at 7amleh-The Arab Center for Social Media Advancement, in a message. “Image recognition, facial recognition, emotional analysis, among other things will increase the power of the surveillance state to violate Palestinian right to privacy and to serve their main goal, which is to create the panopticon feeling among Palestinians that we are being watched all the time, which would make the Palestinian population control easier.”

The educational materials obtained by The Intercept show that Google briefed the Israeli government on using what’s known as sentiment detection, an increasingly controversial and discredited form of machine learning. Google claims that its systems can discern inner feelings from one’s face and statements, a technique commonly rejected as invasive and pseudoscientific, regarded as being little better than phrenology. In June, Microsoft announced that it would no longer offer emotion-detection features through its Azure cloud computing platform — a technology suite comparable to what Google provides with Nimbus — citing the lack of scientific basis.

Google does not appear to share Microsoft’s concerns. One Nimbus presentation touted the “Faces, facial landmarks, emotions”-detection capabilities of Google’s Cloud Vision API, an image analysis toolset. The presentation then offered a demonstration using the enormous grinning face sculpture at the entrance of Sydney’s Luna Park. An included screenshot of the feature ostensibly in action indicates that the massive smiling grin is “very unlikely” to exhibit any of the example emotions. And Google was only able to assess that the famous amusement park is an amusement park with 64 percent certainty, while it guessed that the landmark was a “place of worship” or “Hindu Temple” with 83 percent and 74 percent confidence, respectively.

Google workers who reviewed the documents said they were concerned by their employer’s sale of these technologies to Israel, fearing both their inaccuracy and how they might be used for surveillance or other militarized purposes.

“Vision API is a primary concern to me because it’s so useful for surveillance,” said one worker, who explained that the image analysis would be a natural fit for military and security applications. “Object recognition is useful for targeting, it’s useful for data analysis and data labeling. An AI can comb through collected surveillance feeds in a way a human cannot to find specific people and to identify people, with some error, who look like someone. That’s why these systems are really dangerous.”

The employee — who, like all of the Google workers who spoke to The Intercept, requested anonymity to avoid workplace reprisals — added that they were further alarmed by potential surveillance or other militarized applications of AutoML, another Google AI tool offered through Nimbus. Machine learning is largely the function of training software to recognize patterns in order to make predictions about future observations, for instance by analyzing millions of images of kittens today in order to confidently claim that it’s looking at a photo of a kitten tomorrow. This training process yields what’s known as a “model” — a body of computerized education that can be applied to automatically recognize certain objects and traits in future data.

Training an effective model from scratch is often resource intensive, both financially and computationally. This is not so much of a problem for a world-spanning company like Google, with an unfathomable volume of both money and computing hardware at the ready. Part of Google’s appeal to customers is the option of using a pre-trained model, essentially getting this prediction-making education out of the way and letting customers access a well-trained program that’s benefited from the company’s limitless resources.

“An AI can comb through collected surveillance feeds in a way a human cannot to find specific people and to identify people, with some error, who look like someone. That’s why these systems are really dangerous.”

Cloud Vision is one such pre-trained model, allowing clients to immediately implement a sophisticated prediction system. AutoML, on the other hand, streamlines the process of training a custom-tailored model, using a customer’s own data for a customer’s own designs. Google has placed some limits on Vision — for instance limiting it to face detection, or whether it sees a face, rather than recognition that would identify a person. AutoML, however, would allow Israel to leverage Google’s computing capacity to train new models with its own government data for virtually any purpose it wishes. “Google’s machine learning capabilities along with the Israeli state’s surveillance infrastructure poses a real threat to the human rights of Palestinians,” said Damini Satija, who leads Amnesty International’s Algorithmic Accountability Lab. “The option to use the vast volumes of surveillance data already held by the Israeli government to train the systems only exacerbates these risks.”

Custom models generated through AutoML, one presentation noted, can be downloaded for offline “edge” use — unplugged from the cloud and deployed in the field.

That Nimbus lets Google clients use advanced data analysis and prediction in places and ways that Google has no visibility into creates a risk of abuse, according to Liz O’Sullivan, CEO of the AI auditing startup Parity and a member of the U.S. National Artificial Intelligence Advisory Committee. “Countries can absolutely use AutoML to deploy shoddy surveillance systems that only seem like they work,” O’Sullivan said in a message. “On edge, it’s even worse — think bodycams, traffic cameras, even a handheld device like a phone can become a surveillance machine and Google may not even know it’s happening.”

In one Nimbus webinar reviewed by The Intercept, the potential use and misuse of AutoML was exemplified in a Q&A session following a presentation. An unnamed member of the audience asked the Google Cloud engineers present on the call if it would be possible to process data through Nimbus in order to determine if someone is lying.

“I’m a bit scared to answer that question,” said the engineer conducting the seminar, in an apparent joke. “In principle: Yes. I will expand on it, but the short answer is yes.” Another Google representative then jumped in: “It is possible, assuming that you have the right data, to use the Google infrastructure to train a model to identify how likely it is that a certain person is lying, given the sound of their own voice.” Noting that such a capability would take a tremendous amount of data for the model, the second presenter added that one of the advantages of Nimbus is the ability to tap into Google’s vast computing power to train such a model.

“I’d be very skeptical for the citizens it is meant to protect that these systems can do what is claimed.”

A broad body of research, however, has shown that the very notion of a “lie detector,” whether the simple polygraph or “AI”-based analysis of vocal changes or facial cues, is junk science. While Google’s reps appeared confident that the company could make such a thing possible through sheer computing power, experts in the field say that any attempts to use computers to assess things as profound and intangible as truth and emotion are faulty to the point of danger.

One Google worker who reviewed the documents said they were concerned that the company would even hint at such a scientifically dubious technique. “The answer should have been ‘no,’ because that does not exist,” the worker said. “It seems like it was meant to promote Google technology as powerful, and it’s ultimately really irresponsible to say that when it’s not possible.”

Andrew McStay, a professor of digital media at Bangor University in Wales and head of the Emotional AI Lab, told The Intercept that the lie detector Q&A exchange was “disturbing,” as is Google’s willingness to pitch pseudoscientific AI tools to a national government. “It is [a] wildly divergent field, so any technology built on this is going to automate unreliability,” he said. “Again, those subjected to them will suffer, but I’d be very skeptical for the citizens it is meant to protect that these systems can do what is claimed.”

According to some critics, whether these tools work might be of secondary importance to a company like Google that is eager to tap the ever-lucrative flow of military contract money. Governmental customers too may be willing to suspend disbelief when it comes to promises of vast new techno-powers. “It’s extremely telling that in the webinar PDF that they constantly referred to this as ‘magical AI goodness,’” said Jathan Sadowski, a scholar of automation technologies and research fellow at Monash University, in an interview with The Intercept. “It shows that they’re bullshitting.”

Google, like Microsoft, has its own public list of “AI principles,” a document the company says is an “ethical charter that guides the development and use of artificial intelligence in our research and products.” Among these purported principles is a commitment to not “deploy AI … that cause or are likely to cause overall harm,” including weapons, surveillance, or any application “whose purpose contravenes widely accepted principles of international law and human rights.”

Israel, though, has set up its relationship with Google to shield it from both the company’s principles and any outside scrutiny. Perhaps fearing the fate of the Pentagon’s Project Maven, a Google AI contract felled by intense employee protests, the data centers that power Nimbus will reside on Israeli territory, subject to Israeli law and insulated from political pressures. Last year, the Times of Israel reported that Google would be contractually barred from shutting down Nimbus services or denying access to a particular government office even in response to boycott campaigns.

Google employees interviewed by The Intercept lamented that the company’s AI principles are at best a superficial gesture. “I don’t believe it’s hugely meaningful,” one employee told The Intercept, explaining that the company has interpreted its AI charter so narrowly that it doesn’t apply to companies or governments that buy Google Cloud services. Asked how the AI principles are compatible with the company’s Pentagon work, a Google spokesperson told Defense One, “It means that our technology can be used fairly broadly by the military.”

“Google is backsliding on its commitments to protect people from this kind of misuse of our technology. I am truly afraid for the future of Google and the world.”

Moreover, this employee added that Google lacks both the ability to tell if its principles are being violated and any means of thwarting violations. “Once Google offers these services, we have no technical capacity to monitor what our customers are doing with these services,” the employee said. “They could be doing anything.” Another Google worker told The Intercept, “At a time when already vulnerable populations are facing unprecedented and escalating levels of repression, Google is backsliding on its commitments to protect people from this kind of misuse of our technology. I am truly afraid for the future of Google and the world.”

Ariel Koren, a Google employee who claimed earlier this year that she faced retaliation for raising concerns about Nimbus, said the company’s internal silence on the program continues. “I am deeply concerned that Google has not provided us with any details at all about the scope of the Project Nimbus contract, let alone assuage my concerns of how Google can provide technology to the Israeli government and military (both committing grave human rights abuses against Palestinians daily) while upholding the ethical commitments the company has made to its employees and the public,” she told The Intercept in an email. “I joined Google to promote technology that brings communities together and improves people’s lives, not service a government accused of the crime of apartheid by the world’s two leading human rights organizations.”

Sprawling tech companies have published ethical AI charters to rebut critics who say that their increasingly powerful products are sold unchecked and unsupervised. The same critics often counter that the documents are a form of “ethicswashing” — essentially toothless self-regulatory pledges that provide only the appearance of scruples, pointing to examples like the provisions in Israel’s contract with Google that prevent the company from shutting down its products. “The way that Israel is locking in their service providers through this tender and this contract,” said Sadowski, the Monash University scholar, “I do feel like that is a real innovation in technology procurement.”

To Sadowski, it matters little whether Google believes what it peddles about AI or any other technology. What the company is selling, ultimately, isn’t just software, but power. And whether it’s Israel and the U.S. today or another government tomorrow, Sadowski says that some technologies amplify the exercise of power to such an extent that even their use by a country with a spotless human rights record would provide little reassurance. “Give them these technologies, and see if they don’t get tempted to use them in really evil and awful ways,” he said. “These are not technologies that are just neutral intelligence systems, these are technologies that are ultimately about surveillance, analysis, and control.”

The post Documents Reveal Advanced AI Tools Google Is Selling to Israel appeared first on The Intercept.

This post was originally published on The Intercept.

Amid a historic and ever-worsening humanitarian crisis in Afghanistan, Facebook recently added the head of one of the country’s most important domestic aid groups to its Dangerous Individuals terror blacklist, The Intercept has learned.

Internal company materials reviewed by The Intercept show that Matiul Haq Khalis — head of the Afghan Red Crescent Society, or ARCS; son of a famed mujahedeen commander, Mohammad Yunus Khalis; and a former Taliban negotiator — was added to the company’s stringent censorship list in late April, joining a group of thousands of people and organizations deemed too dangerous to freely discuss or use the platform, including alleged terrorists, hate groups, drug cartels, and mass murderers. But Facebook’s designation now means that the list, ostensibly created and enforced to stop offline harm, could disrupt the work of a globally recognized organization working to ease the immiseration of tens of millions of civilians.

After the collapse of the U.S.-backed government and withdrawal of American military forces, Khalis was named president of the organization, which helps provide health care, food, and other humanitarian aid to civilians there since its founding in 1934. In a country where half the population is going hungry and American sanctions threaten a total economic collapse, the ARCS is a bulwark against even greater suffering. Following Khalis’s addition to the Dangerous Individuals list under its most restrictive “Tier 1” category for terrorists due to his Taliban affiliation, the over 2 billion Facebook and Instagram users around the world are now barred from praising, supporting, or representing Khalis; this means even an anodyne photo of him at an official ARCS event, quotation of remarks, or positive mention of him in the context of the organization’s aid work would risk deletion, as would any attempt on his part to use the company’s platform to communicate, either in Afghanistan or abroad.

“The Afghan Red Crescent continues to provide lifesaving assistance across the country, to the most vulnerable people in the country, working in all provinces,” said Anita Dullard, spokesperson with the International Committee of the Red Cross. “They’re dealing with a range of things including severe drought, Covid, economic hardship, and working to support the healthcare system in Afghanistan. We work closely with Afghan Red Crescent to ensure that we can deliver humanitarian assistance.”

A senior official with a major international aid organization in Afghanistan, who spoke with The Intercept on the condition of anonymity due to avoid jeopardizing operations in the country, described ARCS as “one of the major humanitarian actors delivering services to a growing number of people in need” and “a huge contributor to the collective humanitarian efforts” pursued in conjunction with other NGOs. This aid official expressed surprise that Khalis would be singled out for censorship despite his Taliban affiliation, saying he had “never held a gun,” and expressed concern over the potential to impede lifesaving humanitarian work. “For sure the ARCS is using Facebook as a tool of communication” with the public, this source continued. “If [the blacklisting] has an effect it will be negative” for Afghanistan, they added.

Khalis has had an “extremely varied career” in Afghanistan, according to Graeme Smith, an Afghanistan analyst at the International Crisis Group and former United Nations officer stationed in the country. Smith noted that Khalis was in recent history considered an ally of the U.S., having served with the anti-Soviet mujahedeen led by his father, who in 1987 was feted by President Ronald Reagan at a White House reception. Following the American invasion in 2001, Khalis sided with the Taliban. “In other words he’s from a prominent family with pedigree rooted in tribal support from eastern Afghanistan and a history of fighting invaders,” explained Smith. “I have spent the better part of my career studying Afghan politics and I have never met any important politician who is not ‘dangerous’ in some way. Afghans have learned through bitter experience that Western politicians are also dangerous, at times.”

Facebook’s designation of Khalis, considered in a vacuum, is unsurprising. The company’s Dangerous Organizations and Individuals roster generally mirrors the foreign policy stances of the United States, blacklisting federally sanctioned and terror-designated entities like the Taliban as a matter of course while granting great latitude to Western allies. In Afghanistan, Facebook’s near-total mimicry of State Department decision-making has meant that the ruling government of a sovereign country, as repressive of its own people and despised as it may still be in the U.S., is unable to freely use the internet to communicate with its citizenry. The U.S. government and Facebook share not only a common dilemma over how to treat the Taliban now that the group has won the war and assumed control of the country, but seem to be taking the same punitive approach to that matter. Just as the Biden administration continues to punish the Taliban at the expense of the people of Afghanistan by withholding billions of dollars in frozen cash, Facebook now sanctions the head of one of Afghanistan’s most important humanitarian organizations at a time when Afghans are selling their kidneys to avoid starvation. “It goes without saying that the Red Crescent plays a crucial humanitarian role in Afghanistan’s ongoing armed conflicts,” added Smith.

John Sifton, Asia advocacy director at Human Rights Watch, told The Intercept that he doubted the blacklisting would have a significant impact on relief efforts inside the country, given the relatively small scope of the ARCS compared to larger international organizations. “It’s not going to somehow significantly impact their operations or outreach,” he said. “It’s more illustrative of Facebook having a policy that doesn’t make a lot of sense.” Sifton questioned the extent to which letting people speak freely of Khalis would endanger anyone or anything. “How is he ‘dangerous’? He’s like 65 years old. He has no militia. His father was a mujahedeen commander, but what is the problem here?” Sifton pointed to groups that are actively using the platform to incite violence. “There are hate guys in India that are spreading toxic anti-Muslim violence across Facebook, Hindu nationalist groups, hateful Buddhist groups in Burma, that’s a real problem. Having Khalis online posting about how he cut the ribbon at a new hospital in Afghanistan, that’s not part of the problem.”

Facebook has at times defended the breadth of its blacklist by claiming, without evidence, that it’s legally required to censor discussion of certain entities in order to comply with U.S. sanctions law, though neither the ARCS nor Khalis are currently named in the Treasury or State Department’s counterterrorism sanctions lists. And although the Taliban has an inarguably ugly human rights record and a long history of civilian brutalization, so do many governments left untouched by the Dangerous Organizations policy. The Dangerous Organizations and Individuals list is often criticized for its lack of flexibility and country-specific nuance, and though the company has shown that it is at times willing to make drastic exceptions, these exceptions generally also jibe with American policy determinations.

“The fact that Twitter is doing the exact opposite tells you everything you need to know.”

While Sifton is critical of Facebook’s rigid censorship policies, he also assigns blame to “scattershot” and outdated federal anti-terror policies and dismissed the company’s claims that it has any legal obligation to mimic them: “The fact that Twitter is doing the exact opposite tells you everything you need to know.” Sifton said that by following the “absurdities” of counterterrorism sanctions lists, Facebook is replicating the government’s mistakes. While he emphasized that he was not defending the “misogynist, authoritarian, rights-abusing” Taliban, he questioned the notion that the aging mujahedeen of the 1980s still represent a “danger” to the global community. “The Taliban was dangerous because they hosted Al Qaeda between 1996 and 2001, and Al Qaeda used their territory to plan 9/11 … and all the guys who did that are dead, and all the Arabs they hosted are either dead or very old or at Guantánamo.” To the extent that the Taliban writ large represents a genuine danger to Afghan civilians, it’s unclear how restricting global discussion of Khalis might help.

Facebook did not respond to a request for comment.

Khalis was added to the social network’s blacklist alongside some two dozen other Taliban-affiliated individuals, including others in humanitarian or public health roles, like Afghanistan’s minister of public health, deputy minister of disaster management, and deputy minister of refugees. But unlike these latter offices, the ARCS is nongovernmental, part of the International Red Cross and Red Crescent Movement of humanitarian relief organizations.

In response to a request for comment, the International Federation of Red Cross and Red Crescent Societies provided a statement from ARCS Acting Secretary General Mohammad Nabi Burhan, stating that the Taliban government has not affected the group’s mission or ongoing work. “The Afghan Red Crescent Society delivers impartial, neutral and independent humanitarian services across all provinces in Afghanistan, in its role as auxiliary to public authorities in accordance with the Statutes of the International Red Cross and Red Crescent Movement,” he wrote. “Afghan Red Crescent Society has been operating under a new leadership since October 2021. It is not unusual for changes in leadership of a Red Cross or Red Crescent National Society to follow a change in leadership at a national level.”

The post Facebook Anti-Terror Policy Lands Head of Afghan Red Crescent Society on Censorship List appeared first on The Intercept.

This post was originally published on The Intercept.