According to hundreds of public documents, the FBI and Defense Department were developing face recognition software to identify people using street cameras and drone images. These documents reveal the government’s plans to create a reliable, sophisticated monitoring technology.
The records show FBI, Military, and academic researchers working together to create artificial intelligence tools to identify and track citizens without their consent. After the ACLU filed a Freedom of Information Act complaint, the FBI released these. The Janus initiative was funded by IARPA’s high-level research division, based on the DARPA of the Pentagon. Records frequently mention the project.
According to documents shared by the ACLU with The Washington Post, program leaders worked with FBI scientists and top computer vision specialists. To create and evaluate software that operates swiftly and precisely to identify “really unfettered face imagery” from security cameras in public spaces like subway stations and street corners.
In a 2019 presentation, the IARPA program manager said the goal was to “dramatically improve” facial recognition systems. By “scaling to handle millions of subjects” and instantly recognizing faces from partially occluded views. A system learned to recognize faces at over 0.5 kilometers.
According to the data, researchers paid dozens of volunteers to act out realistic scenarios in a Defense Department training center that resembled a hospital, subway station, outdoor market, and school to improve the system’s capabilities. The test produced thousands of surveillance videos and photos, including drone footage.
The Pentagon’s Combating Terrorism Technical Support Agency, which provides military technology to civil police units, received the upgraded facial recognition technology in Horus, a search engine.
According to Department of Homeland Security officials from a year ago, Horus tool feedback from at least six government departments is still used to improve the program.
Internal emails, presentations, and other documents show how the military and the nation’s top law enforcement agency have actively used a technology that could violate Americans’ privacy and already has a counterpart in pervasive surveillance systems in China, London, Moscow, and other cities.
Even while three states and more than a dozen localities banned or restricted local police use of the technology, the data show that federal officials were more actively involved in its creation than previously thought.
Facial Recognition And Biometric Search Capabilities
Facial recognition technologies have no government regulations. On Tuesday, Sen. Edward J. Markey (D-Mass.) said he would propose a bill to limit government agencies’ use of facial recognition and biometric search capabilities. The 2020 bill was introduced.
Markey told The Post that Americans’ ability to move freely is deteriorating rapidly. We cannot enable the surveillance state to invade our privacy while treating us all as suspects in an unregulated probe that threatens our freedom and liberty.
Nathan Wessler, ACLU deputy director, called domestic mass surveillance a “horror scenario.” “It might allow cops to track anyone for as long as they choose. That’s bad for democracy.
“Committed to deploying facial recognition technology responsibly, ensuring that it appropriately respects individuals’ privacy and civil liberties,” the FBI stated. A Defense Department spokesperson had accepted a comment request by publication, but other inquiries remained unanswered. A spokesman said IARPA is more concerned with technology development than utilization.
After a request and legal battle, the ACLU got the papers in 2019. They don’t specify how the study is being used. Local and federal detectives now employ facial recognition more often.
A 2021 GAO report found that 20 government organizations used facial recognition, including the USPS and FWS. Most firms “had not sufficiently assessed the possible hazards” because they “did not have awareness” of their staff’s equipment.
The documents show how researchers have used artificial intelligence, computer vision, and personal data to make regular technological judgments. In 2019, a top FBI scientist said “social media” and “cell phones with cameras” were the “greatest enablers of better facial recognition” during a Baltimore forensic science luncheon.
Janus, named after the two-faced Roman deity of beginnings and entrances, was formed in 2014 to “radically expand the settings in which automated facial recognition may establish identification,” according to records.
Federal detectives at the time had to use photographs “restricted” from passports or licenses to identify suspects, victims, and witnesses caught on video near a crime scene. These algorithms “severely undervalued and under-exploited all available face information in a video,” according to research.
Research teams developed algorithms to instantly identify and follow the same face across multiple recordings and camera angles to help detectives access new security data. One story said the idea was to make video a benefit.
According to Erik Learned-Miller, a University of Massachusetts at Amherst professor and member of one of the research teams, federal officials involved in the program were careful to distinguish between what They were prepared to sacrifice for American citizens and the capabilities they wished to build for use globally.
The study intended to improve a tool that regional, state, and national law enforcement was using more often. An FBI official told the researchers how the technology would be used to identify child sexual abusers in films, making it hard to deny the purpose.
But, the research’s ambitious goals left him wondering how the findings could be used long-term. He often questioned how the intelligence community would use this knowledge.
The Janus project’s work was consolidated into the Horus web interface, named after the falcon-headed Egyptian god of the heavens, and ended in 2020. “Twice as accurate as the most frequently used government-off-the-shelf technologies,” the Janus initiative enhanced “nearly every aspect of fundamental facial recognition research,” according to IARPA.
The Janus project only partially represents the FBI’s massive facial recognition technical interest. The Interstate Image System searches tens of millions of jail booking photos, scars, and tattoos using a facial recognition search engine. Municipal and state police have this technology.
The FBI’s Next Generation Identity biometric database comprises the photosystem, fingerprints, palm prints, face images, and eye patterns from millions of people who applied for citizenship, were arrested or requested background checks for jobs.
In one of the ACLU’s documents, the Interstate Picture System’s how-to manual notes that it should only be used for “investigative reasons” but that “it is the user agency’s job to determine permissible usage guidelines.”
The forms allow local law enforcement to send suspect photos to the FBI’s Face Analysis, Comparison, and Evaluation (FACE) Services Section, which performs a face recognition search and suggests matches. The form allows officers to request that the images be compared to State Department passport and visa databases and Defense Department biometric databases for foreign nationals and combatants.
In 2019, government auditors reported that the FBI possesses over 640 million face pictures and that the FACE unit performed over 390,000 facial recognition searches in the previous eight years.
Federal agencies have purchased private facial recognition technologies for research. This year, the FBI signed a $120,000 deal with Clearview AI, a facial recognition startup that takes photographs from social media and the internet without consent. Under the contract, FBI officials said they were spending money on “a search engine leveraging publically available photographs… to be used in ways that eventually reduce crime.”
Last year, the Defense Department awarded RealNetworks a $730,000 contract for facial recognition software for “identifying and intelligence-gathering” autonomous drones.
Civil rights groups warn that face recognition research may accelerate US public surveillance. Facial recognition algorithms perform worse when analyzing people of color. They have been implicated in several recent cases of Black males being falsely detained for crimes they did not commit.
Critics have criticized US plans to utilize facial recognition for mass surveillance, like in China. In January, FBI Director Christopher A. Wray expressed alarm over the Chinese government’s development of AI. We could do it. Then, “God, they can.
When the Janus program began, face recognition technology had become “a pervasive, conventional forensic investigative approach,” according to Clare Garvie, a National Association of Criminal Defense Attorneys lawyer who investigated the issue.
Yet, the lack of transparency and doubt about the technology’s reliability in identifying criminal suspects increase the possibility of unfavorable use.
A firm or research organization can claim a technology runs discreetly in a lab. Garvie disagrees that it may be used by drones or defective surveillance to make arrests. Human beta tests for real-world technology are successful.
On Tuesday, an ACLU lawsuit released thousands of records detailing U.S. government involvement in facial recognition technology, sparking new calls for a federal ban. “Americans’ freedom to navigate our communities without continual tracking and surveillance are deteriorating alarmingly speed,” Sen. Ed Markey(D-Mass.) told the Post. “We cannot stand by as the surveillance state digs deeper into our private lives, treating everyone like suspects in an unfettered probe that destroys our rights and freedom.” No federal law restricts facial recognition tools, but some towns and states have. After the study, Markey and other Democrats reintroduced a restriction on government use of the technology.