How AI facial recognition is a threat to us all

An artist’s impression of police upping AI surveillance on Palestine protesters

The police are gloating that the use of facial recognition technology will see convictions of racist rioters. But such technology is a weapon that the state will also use against the left and campaigners. It is an additional element of state repression alongside abhorrent practices, from suppressing protests to the spy cops scandal.

Hiding behind the excuse of far right violence, Keir Starmer called for the “wider deployment of facial recognition technology”. Facial recognition technology (FRT) is powered by Artificial Intelligence (AI). It captures the details of your face and then compares it to a database of pre‑identified faces.

Who is on the database already reflects the state’s view of its enemies and suspects. Live facial recognition technology (LFR) constantly monitors public spaces and captures all the faces it sees. It may be the police on the streets with cameras looking for known suspects, or it may be cameras in shops like M&S looking to defend salmon from known shoplifters.

LFR curtails human rights and can lead to false arrests—all the while helping huge technology companies grab massive profits. Pete Fussey and Daragh Murray published a report in 2019 assessing the London Met police’s trial of LFR. In it they highlight how the problems of LFR go far beyond the obvious privacy violations.

A key issue is that LFR batters rights to freedom of expression and assembly. “The deployment of LFR technology may generate a chilling effect whereby individuals refrain from lawfully exercising their democratic rights due to a fear of the consequences that may follow,” the report says.

People may be afraid to meet with certain individuals and attend certain organisations or protests due to the fear of “guilt by association”. Fussey and Murray’s warning was well-founded. Dr Victor Olisa, a former Met police chief, says that even face coverings and balaclavas are insufficient to remain unidentifiable.

“You might have half a dozen young men and five are masked up and one isn’t. If you can find that one, say on the police database, then you can work your way back,” he said. The report concludes that “police use of overt surveillance technology may impair political protest movements”.

Fussey, who works with the human rights organisation Liberty, has also warned that LFR use will extend beyond far right rioters and fascists that it’s supposedly intended for. “People support more draconian uses of technology when something like this happens. But there’s no meaningful boundary to stop it being used in other contexts,” he said.

Such “function creep”, which sees FRT used on anyone who the state sees as a threat, is a slippery slope. The state’s enemies often aren’t the same as ours.

From history we know the police uses its power to crush left organisations when it claims it wants to stop the far right.

The 1936 Public Order Act was introduced to deal with Oswald Mosley’s fascist. But it was barely used against fascists. Instead it was invoked to stop left organisation. Recent events spark further concern. Police arrested 27 Just Stop Oil (JSO) activists in June on conspiracy charges for planned interference with civil infrastructure.

Extinction Rebellion and JSO co-founder Roger Hallam was sentenced to five years in prison. Already the state is arresting people for planning disturbance and disruption. Now LFR and AI technology provides the police with a more effective method to monitor and attack these organisations.

Police use of LFR to suppress left organisations also widens the scope to use LFR to escalate existing racist policing practices. The institutionally racist nature of the police means LFR will be used to further target black and minority groups.

Unsurprisingly, the first use of LFR was in 2016 at the Notting Hill Carnival in west London. Racial bias is also baked into the technology itself. Because AI is trained and programmed by people and the data they collect, it is always situated within the social biases of that society.

As we live in a racist and sexist society, the AI we produce will reflect those biases—and make them worse. A shocking study from 2018 by Joy Buolamwini and Timnit Gebru, published by MIT Media Lab, showed the error rate for light-skinned men was 0.8 percent, compared to 34.7 percent for darker-skinned women.

Another study by the National Institute of Standards and Technology in the United States found black and Asian people were ten to 100 times more likely to be misidentified than white people. But no laws in Britain have been instituted surrounding the use of FRT, let alone LFR, in law enforcement, unlike in the rest of Europe.

And FRT will also disproportionately target Muslims. The technology cannot differentiate between people wearing face coverings, meaning it could ramp up pre‑existing Islamophobia. Police statistics already show that testing in “ideal conditions” leads to misidentification rates of under 0.02 percent.

But in real-life testing across England and Wales, the rate is nearly 75 percent. This raises questions of what happens to those who are misidentified. In Britain, a teenage girl was misidentified by LFR and wrongly accused of being a shoplifter in a Home Bargains store. She was searched, removed from the store and told she was barred from a number of shops across Britain.

A similar case involved Shaun Thompson, a black social worker, who police held up for nearly half an hour. They demanded his fingerprints because they presumed he was a criminal based on facial recognition. Cases such as these only widen the gap between holding the police to account and letting police treat people however they like.

Who is served by tales of AI’s global ­takeover?

  Read More

The end users and beneficiaries of FRT—the police—should be accountable in cases like the above. But cops brutalise, murder, smash our rights and get away with crime on a daily basis. They already disproportionately stop and search black people and face no consequences.

Arresting people based on harvested facial recognition, whether accurately or not, only provides more scope for abuse from the state and its forces. Rather than address the threat FRT and AI poses to minority groups and democratic rights, police are entering into opaque contracts with private companies. 

The Met police force uses FRT from NEC Corporation, a Japanese-based company that amassed a global revenue of £19.1 billion in 2023. It’s a powerful multinational corporation that has private influence over the police’s use of an unregulated technology.

And as investment into AI continues to increase, its lack of profitability is worrying investors. The New York Times newspaper reported that spending on AI hit nearly £47 billion in April, May and June this year. But returns lie a long way off. That means to reap profits, more of it needs to be sold.

So facial recognition may become a lucrative avenue for huge tech companies, especially if it becomes common police practice in Britain and the US. Meanwhile Starmer has shown his hand. He wants to deploy facial recognition more widely, leaving Labour open to the influence of corporate lobbyists shilling ever-advancing FRT.

Starmer’s commitment to increasing the scope of facial recognition represents his dedication to escalation of police powers—and the state’s clamp down on anyone it doesn’t like. Police use of LFR won’t stop crime or keep people safe. It’s simply another authoritarian avenue for police to suppress our rights.