DUBAI, United Arab Emirates — As a backlash against facial recognition technology grows in the US, a host of Chinese and American firms are competing to supply Dubai’s police force with biometric surveillance and artificial intelligence products.
This month San Francisco banned police from using facial recognition altogether, and other bills in the US aim to do the same. Amazon faced pressure last week from activist shareholders over sales of its Rekognition system to US government authorities. And Rep. Alexandria Ocasio-Cortez expressed concern that face recognition could be used as a form of social control.
But even as the technology comes under more scrutiny in the United States, tech giants such as IBM, and China’s Hikvision and Huawei, are marketing biometric surveillance systems in the UAE, where citizens have fewer options to push back. The UAE has used cellphone hacking software to spy on hundreds of dissidents, journalists, and suspected criminals, and has invested heavily in surveillance technology, according to human rights groups and international media reports.
“It is absolutely terrifying,” said Sarah Aoun, a digital rights technologist who works with human rights activists, including in the Middle East, on privacy and security. “In a place like Dubai, where there is not much freedom of expression and people are being jailed for what they say, when you introduce artificial intelligence, it’s used by systems of power to reinforce their control over the population. They are weaponizing this technology.”
“They are weaponizing this technology.”
Police in Dubai have begun rolling out an ambitious program, dubbed Oyoon, the Arabic word for “eyes,” that will implement facial recognition and analysis driven by artificial intelligence across the city. Police say the program will reduce crime as well as traffic accidents. An analysis of hundreds of government procurement and regulatory documents make clear the scope of Dubai’s high-tech policing ambitions, showing the police have sought video analytics platforms meant to record and analyze people’s faces, voices, behavior, and cars in the time it takes to do a Google search. And a review of dozens of company marketing materials and interviews with officials show global tech giants are eager to provide the police with the technology they are seeking.
Facial recognition is being taken up by police departments and security forces all over the world, from China and Ecuador to the US and UK. Biometric surveillance systems have swept up the faces, voices, and personal traits of millions of people into government databases in countries across the world with little to no transparency or regulatory oversight.
It has sparked an outcry from civil rights advocates who say the technology’s unchecked use in public spaces presents a grave threat to individual privacy and could exacerbate the most abusive practices of police forces, from racial discrimination to arbitrary detention.
At a recent government-organized conference on artificial intelligence in Dubai, representatives from technology companies including Huawei, which the Trump administration recently put on a trade blacklist as a threat to national security, China’s Hikvision, and IBM said they saw the UAE and other countries in the Persian Gulf as an exciting market to sell their video analysis platforms, which they say can do everything from analyzing the behavior of groups to automatically blacklisting individuals based on their faces.
Other governments in the Persian Gulf, such as Bahrain and Saudi Arabia, are also using cellphone hacking and other high-tech surveillance measures to monitor and intimidate dissidents, including exiles. The UAE meanwhile has poured money into developing its surveillance capabilities.
“In looking at the way they’re using spyware, they are more aggressive about becoming a surveillance power,” said Bill Marczak, a postdoctoral researcher at the University of California, Berkeley, and a research fellow at Citizen Lab who is one of the foremost analysts of surveillance systems in the Persian Gulf. “They’re not only targeting people in the UAE but they’re also using spyware beyond their borders to conduct operations in other countries.”
The police have said little about the specific kinds of technologies that are being used in the Oyoon program. BuzzFeed News reviewed procurement records that show the police have sought face and voice recognition technology since as early as mid-2016, including products like the Vectra M3, a program marketed toward skincare professionals that aims to create a 3D model of a person’s face, and NeoFace Reveal, a software made by the Japanese company NEC that says it matches faces from photographs and crime scene videos to preexisting databases of images. The police also sought a voice recognition program called Batvox, described as a “new public security weapon” by its manufacturer, Massachusetts-based Nuance.
The Oyoon program, which was unveiled last January, aims to bring thousands of video feeds from cameras used by government authorities throughout Dubai into a central command center, according to local news reports and people familiar with the program. The command center uses artificial intelligence and face recognition to identify security threats. “[Oyoon] security system employs artificial intelligence to watch Dubai around the clock!” proclaims a graphic on a PowerPoint slide in a presentation on the UAE’s artificial intelligence strategy given by an adviser to the UAE government.
Police in Dubai have boasted the program has already led to the arrest of more than 300 people.
Studies have shown racial bias baked into leading facial recognition platforms, raising questions about misidentification of suspects. In the UAE and elsewhere, there have been no independent audits of these platforms for accuracy or freedom from bias.
“The application of artificial intelligence to CCTV systems opens a Pandora’s Box of uses and misuses,” said Sam Samuels, a former police officer and consultant on CCTV systems in Dubai. “At the end of the day it will be a human being who has to validate the information generated by the system.”
While facial recognition has proved divisive in the US, it is in China that its use has been particularly controversial. The Chinese government has come under fire internationally for large-scale surveillance of ethnic minorities. Hikvision, the world’s largest supplier of surveillance products, already supplies Dubai — as well as cities from London to Shanghai — with thousands of CCTV cameras. It has also outfitted mosques and internment camps with face recognition cameras in China’s far west — a region where Muslim minorities are under unprecedented surveillance and where upward of a million people have disappeared into the camps. But at the AI conference in Dubai, Robert Wang, an executive for Hikvision in the Middle East, did not mention the company’s work in Xinjiang. Instead, he set up China’s approach to surveillance and security as an example for the audience of UAE government officials and corporate executives.
“The challenge for us now is not that we no longer have enough video. The problem is that we have too many videos.”
China, he said, is one of the safest countries in the world, saying it needs relatively few police officers per capita.
He said the reason is the 250 million surveillance cameras the country’s domestic security authority has installed throughout the country.
“This kind of massive data provides for us all of the evidence, and all of the clues,” Wang said. “The challenge for us now is not that we no longer have enough video. The problem is that we have too many videos.”
BuzzFeed News spoke to several former detainees, prisoners, and human rights advocates who focus on the UAE. They said the widespread use of face recognition in Dubai is likely to lead to misuse and could do serious damage to privacy rights.
“This is good technology if it’s used in the right way,” said Khaled Ahmad, a computer engineer, prisoners’ rights campaigner, and former detainee in the UAE. “But with the UAE government, they arrest people just for tweeting, for publishing on Facebook, just for speaking about freedom and human rights. It’s good technology if it’s used against criminals, but not if you use it to cover people’s mouths.”
The Oyoon program is being rolled out in cooperation with several UAE government ministries, including the State Security Agency, according to the Dubai police. In a sign of how seriously Dubai’s leaders are taking the program, Dubai’s Crown Prince Sheikh Hamdan bin Mohammed bin Rashid Al Maktoum visited a local police station in a district where facial recognition–equipped cameras were already thought to be in place last October.
The police also sought programs to help with video analytics, extracting deleted data from computers, recognizing license plates, and breaking into cellphones, procurement records showed.
But the true power of video analytics lies in not any one of these functions, but in the profiles it enables law enforcement to build of individuals when many sources of data are put together.
Dubai police did not reply to repeated requests for comment on the Oyoon program by the time of publication.
Hikvision and Huawei declined to comment on the statements in this article.
A spokesperson for IBM said the company has “robust processes in place to ensure potential client engagements are consistent with our values, as well as US and local laws.”
Facial recognition has been around for more than a decade. One of its earliest known uses for law enforcement was in Tampa, Florida, in 2001. The technology remains problematic and has faced questions about everything from its ability to distinguish between nonwhite faces to its overall accuracy, but it is now being widely used by law enforcement agencies to build profiles of individuals in new ways.
File storage costs have become significantly cheaper over the past few years, making it much easier for users to store video. And algorithms that adjust to compensate for real-world variations in people’s appearances, as well as new technology that makes it easy to slice videos frame by frame, have made facial recognition and other biometric surveillance programs cheaper and more effective, said Os Keyes, a PhD student at the University of Washington who is researching the history of facial recognition.
One thing that’s changed is the speed at which sophisticated computer programs quickly search and analyze thousands of hours of camera footage. A camera in a shopping mall, for instance, can capture many different angles of your face over time, enabling a computer to model the way your head looks from different angles and making it easier for police to spot you.
“What they’ve come up with is algorithms that, instead of comparing two photos of you, take a whole sequence of photos of you and use it to construct a 3D model of your head, which they can then rotate to the angle of a photo they already have,” Keyes said. “If you’re taking a video, you can just get 10 frames of a person and wait for one frame where they look up and don’t have a baseball cap in the way.”
And now video analytics programs can search footage for combinations of people’s traits. An official from BriefCam, an Israeli firm that is owned by Japan’s Canon and has a tech partnership with HP Enterprise, demonstrated this at the AI conference in Dubai, showing a screen with a search bar that pulled up dozens of women who had worn red dresses with midlength sleeves in a shopping mall.
“The main use case is security,” the BriefCam official said, adding that BriefCam would be happy for the opportunity to work with the UAE security apparatus.
Stephanie Weagle, chief marketing officer for BriefCam, said in a statement that biometric technology has “significant public safety benefits” including identifying suspects and finding missing children. “The expectation is that technology designed for public safety is used in accordance with applicable laws and to uphold civil rights,” she said.
China’s Huawei displayed a data analysis platform labeled FusionInsight that could quickly comb through thousands of hours of footage for faces and license plates. A company official said the platform had the ability to blacklist individuals. When asked about this, Huawei declined to comment.
Microsoft displayed a video analytics platform that claims to detect the face, age, and emotional state of people seen on camera, as well as the size of crowds, though company officials did not refer to applications in law enforcement. Asked whether the platform is being used in the UAE, a spokesperson for Microsoft pointed BuzzFeed News to a blog post by the company’s president calling for governments to regulate facial recognition technology.
To illustrate the prowess of Hikvision’s technology, Wang, the company executive, showed a six-minute film shot in the style of a primetime cop drama. In the film, a team of Chinese police officers, whose voices had been dubbed in American-accented English, use the company’s suite of products to track down a jewel thief wearing a black ski mask. The police eventually find the suspect using footage of a small black tattoo on his wrist, matching images of his license plate–less car, searching for people who frequently appeared near the jewelry shop in the weeks past, and directing cameras in public spaces to find him by using images of his face. The man later tries to hide from cameras by wearing a baseball cap, but it doesn’t work.
“Add him to the criminal database and track him down,” the lead police officer barks.
At the end of the film, Wang told the audience the surveillance technology, and the way it could work together, was all real — “not science fiction,” he said.
“This is not a system that one manufacturer can build alone,” he added. “I’m glad to be here today if users and partners are interested. We would like to be open and to cooperate.”
Chinese companies like Hikvision, which have benefited from Chinese government programs promoting the development of homegrown artificial intelligence technologies and from being able to test their tech on the Chinese public, have become leading players in facial recognition. Hong Kong–based SenseTime, the world’s most valuable AI startup, told BuzzFeed News it is setting up an office in Dubai — the company’s first outside Asia. Discussions between Smart Dubai, the city’s “smart city” authority, and SenseTime about implementation of its facial recognition products are progressing, according to people with knowledge of the matter. Smart Dubai said it met with SenseTime in Shanghai last year to identify possibilities for working together. Asked about the discussions, SenseTime said it could not disclose “commercially sensitive information.”
Hong Kong–based SenseTime, the world’s most valuable AI startup, told BuzzFeed News it is setting up an office in Dubai — the company’s first outside Asia.
George Huang, head of SenseTime’s international business group, said at the conference that mining data from social media could help provide information for public security purposes.
The data helps SenseTime and other AI companies around the world train algorithms to manage public safety, he said. Data like this, he added, is crucial for smart cities.
A spokesperson for SenseTime said the company is “committed to fair and responsible applications of AI technology. In addition to complying with local laws and regulations in the markets we operate in, we also have internal processes to safeguard our technology to ensure it is being used in a fair and responsible manner.”
Beyond recognizing individuals, video analytics platforms have the ability to alert law enforcement authorities to gatherings of people in real time. IBM, Huawei, and other companies displayed platforms that could automatically measure both crowd sizes and characteristics — for example, flagging a certain article of clothing, whether someone was walking in a certain direction, or whether someone was lingering for too long. It’s easy to imagine uses for this kind of technology outside policing — determining whether everyone on a dangerous construction site is wearing a hard hat, for instance, or automatically flagging an injured person who has keeled over. But video analytics platforms like these can also easily be used by security services to quickly crush protests and other peaceful gatherings. A representative for IBM at the conference was quick to tout the company’s PowerAI Vision platform’s utility for law enforcement, saying the company would like to sell it to police and security services in the UAE. A spokesperson for IBM noted the platform could not recognize individual faces, but only label “objects” in a video. Demonstrations of the product show labels on both objects and human beings.
“Watching the traffic of people, we can prevent crimes and terrorism.”
Huang promoted SenseTime’s crowd analysis software as key to managing the density of crowds. It can automatically count the number of people present and analyze the level of safety, he said, flagging “abnormal behavior,” including too many people gathered at once. “Watching the traffic of people, we can prevent crimes and terrorism,” he said.
“We can understand the time, the location, the method of the crime, the moving habits and area of activity of the suspect,” he added. “That allows the city to invest more efficiently in fighting criminals.”
A demo of SenseTime’s product showed a crowd with color-coded circles that appeared around people’s bodies, when they loitered too long, for instance, or when “chaos” happened, referring to disorder.
The extent to which Dubai’s security services and law enforcement already use facial recognition cameras — as well as complementary tech like gait recognition, car recognition, and data analysis — is not known, though analysts said it is likely widely in use. Requirements published by Dubai’s regulator for the security industry specify that private video surveillance systems must be connected to a centralized system called VideoGuard, which allows Dubai police system access.
“Every single conference center, every single hotel, every single road today has a camera. That is true of the UAE, it’s true in the US, it’s true in China and everywhere,” Omar Al Olama, UAE minister of state for artificial intelligence, told a Gulf News reporter in May. “We have no ambition to do this for the sake of surveillance. We’re going to do it if there is a way for us to become the safest city on earth.”
Dubai police have purchased technology for license plate detection, which developed much earlier than facial recognition, for several years, according to procurement records. And as part of the Oyoon program, Dubai has placed facial recognition cameras in at least one city neighborhood called Al Muraqqabat. Located in eastern Dubai, it’s a relatively dense residential area that’s also home to shopping malls and cinemas. Industry analysts said it is likely that face recognition cameras and other biometric surveillance systems are already in wider use in the UAE, including in Dubai and Abu Dhabi, but that authorities have not yet made it public.
The UAE, with a population of about 9.5 million people, has among the highest rates of political prisoners per capita in the world, according to scholar Kristian Coates Ulrichsen. In the years since the Arab Spring uprisings swept across the Middle East, the UAE has invested in surveillance technology, which it has used to spy on both activists and dissenters within its own borders, as well as conduct espionage operations abroad. Reuters reported in January that the UAE retained US operatives to hack into the iPhones of hundreds of activists, political leaders, and suspected terrorists.
“They focus on preventative surveillance,” said Joe Odell, a campaigner at the International Campaign for Freedom in the UAE. “It’s about control to prevent street mobilizations through establishing a wide-reaching surveillance state, where they can nip anything in the bud before it even happens. They’ve spent millions of pounds on that.”
Human rights activists say torture is common in the prison system. Several former UAE prisoners interviewed by BuzzFeed News, who had been jailed or detained over allegations ranging from leaking information to defaming the country, said they had faced abuses and torture in prison, including beatings, electric shocks, and sleep deprivation. They also said they had their eyes scanned, their fingerprints taken, and their devices confiscated — treatment that prisoners in countries including the US also face. Asked about these allegations, Dubai police said the issue was not within their jurisdiction.
Ahmad, the Palestinian computer engineer, did technical support for a firm controlled by a senior Emirati government official before being detained without explanation in Abu Dhabi for months. He said he was blindfolded and taken to a secret prison whose location he still does not know. During interrogations about his work — police seemed to believe he had access to sensitive information — security agents put him in an electric chair, turned it on for 20 seconds at a time, and threw cold water on his face to revive him when he passed out from the pain.
Ahmad said one of the first things the police did was confiscate his cellphone and computer. They demanded the passwords, and he gave them incorrect ones, hoping to stall for time. His interrogators later told him they had broken into the devices anyway. He believed them, he said — one day an officer began joking with him about a woman friend whose pictures were on his phone. She was sexy, the officer said.
Five months after he was first detained, Ahmad was blindfolded and taken by car back to the airport in Abu Dhabi. He was deported to Lebanon without any explanation as to why he was detained in the first place. The police even apologized and told him he was innocent, Ahmad said.
“They’ve imprisoned all those people or intimidated and harassed them into keeping quiet.”
Ahmad eventually moved to Sweden, where he now lives; after he became a Swedish citizen, he began writing about his ordeal in prison. His former employer sent him a letter, which he shared with BuzzFeed News, stating that he could be prosecuted under the UAE’s cybercrime law for allegedly stealing confidential information from the company. Though he now lives abroad and says he has nothing to hide, the idea that he’s being surveilled through his devices is something he thinks about all the time.
Last year, Ahmed Mansoor, who is perhaps the country’s best known human rights activist, was sentenced to 10 years in prison and fined about $270,000 over posts he made on Facebook and Twitter that a court found had damaged the UAE’s reputation abroad as well as its “social harmony and unity.” People in the UAE can be arrested under the country’s cybercrime law, which bans defamatory statements on social media.
Mansoor, who had called attention to human rights abuses throughout the Middle East on his Twitter account, had been under heavy surveillance by his government and had been targeted with spyware on his devices.
“Ahmed was very open about what he did and why he did it, but he was very conscious of being under surveillance. He would meet people in hotel lobbies, and he’d leave his phone at home,” said Nicholas McGeehan, a researcher who formerly worked for Human Rights Watch in the region. “He assumed he’d be under surveillance anyway. You couldn’t make a phone call, couldn’t send an email — everything was funneled through one or two interlocutors who were willing to take a chance.”
“They’ve imprisoned all those people or intimidated and harassed them into keeping quiet,” he added.
Controls on social media and the perception that authorities may be spying on cellphones has already chilled discourse, say free speech advocates and former UAE prisoners. Rights groups said they worried the growing presence of surveillance technology in physical spaces could make things even harder.
“The government has shown a willingness and a desire to go after peaceful opposition,” said Citizen Lab’s Marczak, who has worked with Mansoor. “If they have powerful tools to track your movements outside your house, and then they have spyware to hack the devices inside your house, then there’s pretty much nowhere that you’re not being watched.” ●
Alexandria Ocasio-Cortez’s name was misspelled in an earlier version of this post.