Your Instagram pictures could be part of a facial recognition database thats been made available to law enforcement agencies. Thats thanks to Clearview AI, a mysterious startup that has scraped billions of images from across the web, including from social media platforms like Instagram and Twitter.
Law enforcement has been using facial recognition for a while. But Clearviews technology represents a scary step further than anything weve seen before, according to reporting from the New York Times. The secretive company says its created a database of over 3 billion images that have been scraped from all corners of the internet, including social networks like Facebook, Instagram, and YouTube. From just a snapshot or video still, Clearview claims its app lets a police officer identify a face and match it with publicly-available information about the person, within just a few seconds.
But is this a world we want to live in? Clearview argues that the tech can help track down dangerous people its site points to child molesters, murderers, suspected terrorists and is only meant for use by law enforcement. As the Times reported last week, the companys facial recognition has helped identify child victims in exploitative videos posted to the web. But critics say the technology is way too risky, enabling excessive surveillance and threatening our privacy rights. Another concern is that facial recognition, broadly, has also been shown to be less accurate on people of color, women, and other minority groups.
Faced with these concerns, the worlds biggest tech companies are stepping up, sending cease-and-desist letters to Clearview that order the company to stop scraping their sites for our data. But its not clear how much power those companies have, or how invested they actually are in protecting our personal information. While some lawsuits against Clearview are also popping up, its not yet apparent how Clearview could be stopped. That has privacy advocates pointing to the need for a federal law regulating, or even outright banning, facial recognition in the United States.
Facial recognition isnt new. But this huge database of faces is.
So heres how Clearviews tool works. Say you have an image of a person, but you dont know their name. You could input that photo into Clearviews app, and it will turn up any image of the person that it had scraped from the internet, as well as links to websites from which those images came. That could be a good amount of information.
Again, Clearviews database reportedly includes more than 3 billion images taken from around the web. Thats much more than what law enforcement agencies typically have access to. The Times reports that the technology will work with images of faces from many different angles, while older facial recognition tools used by police departments might require the subject to be looking straight ahead, like in a mug shot.
That means images from social media posts on platforms like Instagram could pop up even images that are no longer, but once were, publicly available. And keep in mind: the tool doesnt just surface pictures that youve taken and posted online. It will also turn up any photos posted of you, even those posted without your consent or knowledge.
Clearview is a search engine for publicly available images, Clearview CEO Hoan Ton-That told Recode in an email. We do not and cannot index any images that are private or protected, such as those in a private Instagram account. A previously public image may or may not be searchable with Clearview depending on the time it was public and that would depend on the individual case.
Clearview decides who can and cant use this tool
More than 600 law enforcement agencies have used Clearview AI in the past year, as have federal agencies like the Federal Bureau of Investigation (FBI) and the Department of Homeland Security (DHS), according to the Times. The FBI would not confirm to Recode that it had used Clearviews tool, instead pointing to its testimony last year about its use of facial recognition more broadly from June 2019. DHS did not respond to Recodes request for comment by the time of publication.
The tool has also been provided to some companies for security, though Ton-That wouldnt tell Recode which ones. The Times has reported that at least a handful of companies have obtained licenses for Clearviews technology for security purposes. Meanwhile, Ton-That told Recode: We decline to comment due to confidentiality and other reasons.
Its also unclear who else, including foreign governments, Clearview is willing to do business with. Ton-That told the Times, If its a country where its just governed terribly, or whatever, I dont know if wed feel comfortable, you know, selling to certain countries. Nevertheless, BuzzFeed News recently reported that Clearview has plans to sell the tool to at least 22 countries, including some with concerning human rights records. Clearview told Recode that it did not currently have contracts outside the US and Canada. The Royal Canadian Mounted Police, the national police service in Canada, told Recode that it does not comment on specific investigative tools but that it researches emerging technologies.
The use of this tool by law enforcement alone raises questions, however. Imagine, for instance, the ethical problems at play with a police officer using Clearview to identify a protester. Or, say the facial recognition doesnt work as it should, and, a false match ultimately leads to arresting someone for a crime they didnt commit.
But theres also fear that Clearviews technology could one day be made available to anyone, and such a development could destroy our expectation of being anonymous in public. Its not difficult to imagine terrifying uses of this. Imagine if a nude picture of you was, at some point in time, posted online. With the snap of a phone camera, its possible that anyone with Clearview could instantaneously find that image. Or imagine youre walking down the street, and someone decides they want to know where you live and whether you had kids. All they might need is the Clearview app.
The scope of Clearviews threat to privacy remains unclear. The Times reported: Police officers and Clearviews investors predict that its app will eventually be available to the public. Thats different from what Ton-That told Recode.
Our strict commitment at this time and at all times previously is and has been not to make this tool available to the general public, Ton-That said in an email. Our mission is to reduce crime, fraud, and abuse using our powerful new technology. Any abuse of our technology would be in total violation of our mission and values.
There are some good reasons to not trust the company
Clearviews past has raised alarm bells. For one thing, CEO Hoan Ton-Thats previous ventures included an app that added Trumps hair onto photos of people, and hes also been linked to a phishingsite. At one point, the company tried to sell a database for extreme opposition research to Paul Nehlen, a white supremacist and anti-Semite, who was running for Congress at the time, according to the Times. One investor in the company is Peter Thiel, who helped found PayPay and Palantir and has vocally supported President Trump.
The company hasnt been exactly forthright in its advertising claims, either. For instance, BuzzFeed News reports found two cases in which Clearview claimed a certain law enforcement agency was using its product, when in fact the agency had simply received a tip from the company. Clearview also appears to be claiming that hundreds of police departments were working with them when some of those police departments only signed up for a trial, according to BuzzFeed News. Ton-That told Recode, Our calculations are based on the number of agencies that are actively using Clearviews technology to help solve crimes.
Its not quite clear how well the tool actually works. According to marketing materials obtained by BuzzFeed News, Clearview claimed to have accuracy finding a match out of 1 million faces 98.6 percent of the time. At the same time, Clearview told the New York Times that the tool produces a match up to 75 percent of the time, although we dont how many of those are true matches. Ton-That also told the Times that one difficulty, for Clearviews algorithms, is that photos scraped from the web were generally at eye-level, which is not at the same angle typically captured by surveillance cameras.
The company had also said that it had checked the accuracy of its tool using a methodology from the American Civil Liberties Union (ACLU), an assertion the organization has pushed strongly back against. The claim has since been removed from Clearviews site, though the company still says that its tool was reviewed and certified by an independent panel of experts. (Recode asked Clearview who those experts were, and well update this post if we hear back).
Meanwhile, Craig Watson, the director of the National Institute of Standard and Technologys Image group, said that Clearview had not volunteered to participate in its facial recognition vendor testing program, and there were no plans to evaluate its algorithms.
Tech companies are fighting back, but that might not be enough
Major internet platforms have responded to Clearview by sending the company cease-and-desist letters. These companies now include LinkedIn, Twitter, Facebook, Venmo, Google, and YouTube. Its worth noting that Twitter explicitly bans using its platform for the purpose of facial recognition. Meanwhile, Pinterest told Recode that it wasnt aware of scraping on its site by Clearview, though such web-scraping would violate their community guidelines. When asked about Clearview scraping its images and videos, PornHub Vice President Blake White said in a statement, We have no knowledge of this and actively block scrapers as they violate our Terms of Service.
Still, none of the tech companies that responded to Recodes request for comment said whether they planned to escalate their demands with lawsuits. Clearview said that its received these letters from companies, and that its attorneys are responding appropriately. In an interview with CBS, the companys CEO insisted that there is a First Amendment right to public information.
Facial recognition software has become increasingly complex and accurate in recent years.
David McNew/AFP via Getty Images
As users of these technology platforms, members of the public are now in a curious position. Although weve long criticized these platforms for profiting off our data, were now potentially reliant on these companies to defend us from a dystopian world of facial recognition. Keep in mind that several of these companies, like Google and Facebook, are have worked on facial recognition technology of their own, albeit confronted with varying levels of controversy and ethical concerns. Just last month, we learned that Facebook had agreed to pay $550 million to settle an Illinois facial recognition lawsuit inspired by its photo-tagging suggestion feature. But despite having access to billions of our photos, none of these companies have gone forward with creating such a tool a boundary Clearview has now crossed.
So Clearview might have your photos. Can you do anything about it?
According to the privacy policy listed on Clearviews site, users have certain rights. But the policy says that the right to erasure deleting your information from their platform is only available under certain conditions. Although he would not say how many, Ton-That told Recode that Clearview has received a number of these requests and is responding appropriately.
In a later email, Ton-That said: Were processing removal requests for persons in jurisdictions that impose that legal requirement. In order to make those requests, you would need to confirm your identity with Clearview by sending in get this a photo of yourself on a government ID.
But if you (understandably) dont feel comfortable doing that, you might consider suing Clearview. At least two lawsuits have already been filed in Illinois, and another lawsuit has been filed against Clearview in Virginia. While Texas also has a biometric privacy law, it requires the states attorney general to take action. That hasnt happened yet.
Ton-That said that Clearview designed its tool to follow all relevant laws and regulation.
Scott Drury, a former Illinois state representative and attorney representing one of those class action lawsuits, told Recode theyre not just arguing that Clearview violated the Illinois Biometric Privacy Act. He said theyre also suing on constitutional grounds.
In this case, you have citizens who clearly have a right to their privacy and they have a right to know how the photographs that they put online are being used, Drury said. And Clearview, working with law enforcement, specifically and covertly took these photos and scraped them from the internet and didnt let anyone know that they were doing that.
The controversy has prompted some police departments to speak out. Now, the NYPD is denying that it has an institutional relationship with the company, according to BuzzFeed News. The Chicago Police Department has said that, after a trial, it spent nearly $50,000 to use the tech for two years, according to the Chicago Sun-Times. Meanwhile, the state attorney general of New Jersey has ordered that all police in the state stop using the tool.
A few cities have already banned law enforcement from using facial recognition. The revelations about Clearview AI have also bolstered calls for federal regulation of the technology.
Congress needs to stop messing around and pass legislation to ban the use of face surveillance technology nationwide, said Evan Greer, the deputy director for the digital rights group Fight for the Future, in a statement. That mimics what other privacy advocates are saying.
And some lawmakers appear to be listening. When the New York Times story came out, Sen. Cory Booker also tweeted that if Congress doesnt act to limit the use of technology in this manner we risk losing our privacy and our freedom. And at the end of January, Sen. Ed Markey wrote to Ton-That demanding a list of all law enforcement and intelligence agencies in communication with Clearview, along with other questions. His office is continuing to look at the company, and says it will be taking further steps.
Still, facial recognition has long been used by law enforcement agencies. So while its possible that Clearview AI could finally galvanize enough backlash to get lawmakers to act, only time will tell. In the meantime, now would be a great time to switch your social media accounts to private. You friends might care what you were wearing Saturday night, but maybe thats something the cops dont need to know.
Open Sourced is made possible by the Omidyar Network. All Open Sourced content is editorially independent and produced by our journalists.