America's top maker of cop body cameras says facial-recog AI isn't safe

You listening, Cressida Dick?


Analysis America's largest manufacturer of body cameras – and the biggest supplier to police forces across the United States – says today's facial recognition technology is not safe for making serious decisions.

Speaking during its second-quarter earnings call with investors this week, the CEO of Axon, Rick Smith, answered a question about whether the company would be adding facial-recognition systems to its suite of products and, if so, whether that would come with an additional cost.

Smith responded in clear terms that current facial recognition is simply not accurate enough to "make operational decisions," ie: for police to use it to recognize individuals and use positive responses as justification for automatically and unquestioningly apprehending people. Well, the computer says you're wanted, so here come the cuffs, we can imagine a conversation with officers going.

"We don’t have a timeline to launch facial recognition," Smith said on the conference call (listen in at around the 40-minute mark), noting that Axon doesn't have a team "actively developing it" either. He added: "This is technology that we don’t believe the accuracy thresholds are right where they need to be to make operational decision off of facial recognition."

That blunt assessment is important due to increasingly widespread use of the technology both by social media companies like Facebook, smart home companies like Nest, and police forces like London's Metropolitan police.

There is a real risk that because facial recognition can be seen to be working in closely defined situations – Facebook and Nest can, for example, run facial recognition against a small set of likely individuals: your friends or people that have visited your house – that more serious applications are considered before the technology is ready.

Regulation

Last month, Microsoft called for Congress to regulate the US government's use of facial-recognition technology – technology that it some cases it provides – after it came under fire for selling face-probing technology to Uncle Sam's Immigration and Customs Enforcement (ICE) agency.

London, UK - March, 2018. Police officers patrolling Leicester Square and Piccadilly Circus in central London. Pic Paolo Paradiso / Shutterstock.com

Zero arrests, 2 correct matches, no criminals: London cops' facial recog tech slammed

READ MORE

The Met Police, over in London, England, has been also been running a controversial trial of facial recognition software at a number of public events – something that led its head, Cressida Dick, being quizzed by the London Assembly.

During that grilling, Dick acknowledged the trials had not been very successful: reports showed a 98 per cent false-positive rate. "It’s a tool, it’s a tactic. I’m not expecting it to result in lots of arrests," she said – but admitted the Met intends to keep plodding ahead with it regardless.

And that is in addition the UK Home Office asking for tenders for a £4.6m ($5.9m) contract to build a facial recognition system and database. The UK police's drive to introduce the technology despite its terrible performance and privacy concerns, even invited the ire of the UK Biometrics Commissioner, Professor Paul Wiles, who criticized the government for failing to produce a formal document outlining its biometrics strategy, and argued that it needs to provide a proper public accounting of such public trials.

Wiles is also concerned about an estimated £100m that has been plowed into efforts to link up cops’ IT systems and develop digital forensics and in his annual report noted that the police’s use of new biometric tech isn’t always organized or systemic, with a "worrying vacuum" in governance and lack of oversight.

The long-delayed biometrics strategy from the UK government finalized emerged this summer, runs to a puny 14 pages, and was described looking like "a late piece of homework with a remarkable lack of any strategy.”

Threshold

Back in the United States, while Axon's CEO was carefully to stress his company is not planning to offer facial recognition any time soon, he also made plain that he felt it was coming and the company would offer it as soon as it met the necessary "accuracy thresholds."

He also warned that the entire technology could be "imperiled" if it was rolled out too early. "Once we’ve got a tight understanding of the privacy and accountability controls, we need to ensure that it will be acceptable by the public at large," he argued. "At that point we would move into commercialization of that capability."

Smith went on: "But this is one where we think you don’t want to be premature and end up with technical failures with disastrous outcomes, or something where there’s some unintended use case where it ends up being unacceptable publicly, and imperils the long term use of the technology." ®

Hat-tip to Dave Gershgorn of Quartz for, from we can tell, first noting Axon's comments.

Similar topics


Other stories you might like

  • Clearview AI promises not to sell face-recognition database to most US businesses
    Caveats apply, your privacy may vary

    Clearview AI has promised to stop selling its controversial face-recognizing tech to most private US companies in a settlement proposed this week with the ACLU.

    The New-York-based startup made headlines in 2020 for scraping billions of images from people's public social media pages. These photographs were used to build a facial-recognition database system, allowing the biz to link future snaps of people to their past and current online profiles.

    Clearview's software can, for example, be shown a face from a CCTV still, and if it recognizes the person from its database, it can return not only the URLs to that person's social networking pages, from where they were first seen, but also copies that allow that person to be identified, traced, and contacted.

    Continue reading
  • Research finds data poisoning can't defeat facial recognition
    Someone can just code an antidote and you're back to square one

    If there was ever a reason to think data poisoning could fool facial-recognition software, a recently published paper showed that reasoning is bunk.

    Data poisoning software alters images by manipulating individual pixels to trick machine-learning systems. These changes are invisible to the naked eye, but if effective they make the tweaked pictures useless to facial-recognition tools – whatever is in the image can't be recognized. This could be useful for photos uploaded to the web, for example, to avoid recognition. It turns out, this code may not be that effective.

    Researchers at Stanford University, Oregon State University, and Google teamed up for a paper in which they single out two particular reasons why data poisoning won't keep people safe. First, the applications written to "poison" photographs are typically freely available online and can be studied to find ways to defeat them. Second, there's no reason to assume a poisoned photo will be effective against future recognition models.

    Continue reading
  • 1,000-plus AI-generated LinkedIn faces uncovered
    More than 70 businesses created fake profiles to close sales

    Two Stanford researchers have fallen down a LinkedIn rabbit hole, finding over 1,000 fake profiles using AI-generated faces at the bottom.

    Renée DiResta and Josh Goldstein from the Stanford Internet Observatory made the discovery after DiResta was messaged by a profile reported to belong to a "Keenan Ramsey". It looked like a normal software sales pitch at first glance, but upon further investigation, it became apparent that Ramsey was an entirely fictitious person.

    While the picture appeared to be a standard corporate headshot, it also included multiple red flags that point to it being an AI-generated face like those generated by websites like This Person Does Not Exist. DiResta was specifically tipped off by the alignment of Ramsey's eyes (the dead center of the photo), her earrings (she was only wearing one) and her hair, several bits of which blurred into the background. 

    Continue reading
  • Face Off: IRS kills plan to verify taxpayers with facial recognition database
    Uncle Sam takes security, privacy concerns seriously, it says here

    Updated The Internal Revenue Service has abandoned its plan to verify the identities of US taxpayers using a private contractor's facial recognition technology after both Democrats and Republicans actively opposed the deal.

    US Senator Ron Wyden (D-OR) on Monday said Treasury Department officials informed his office that the agency has decided to move away from using the private facial recognition service ID.me to verify IRS.gov accounts.

    "The Treasury Department has made the smart decision to direct the IRS to transition away from using the controversial ID.me verification service, as I requested earlier today," Wyden said in a statement. "I understand the transition process may take time, but I appreciate that the administration recognizes that privacy and security are not mutually exclusive and no one should be forced to submit to facial recognition to access critical government services."

    Continue reading
  • Ukraine uses Clearview AI facial-recognition technology
    Controversial search engine being used to identify dead and Russian operatives

    The Ukrainian government is using facial recognition technology from startup Clearview AI to help them identify the dead, reveal Russian assailants, and combat misinformation from the Russian government and its allies.

    Reuters reported yesterday that the country's Ministry of Defense began using Clearview's search engine for faces over the weekend.

    The vendor offered free access to the search engine, which Ukraine is using for such tasks as identifying people of interest at checkpoints and identifying people killed during Russia's invasion, the news organization wrote, citing Lee Wolosky, who currently advises Clearview and formerly worked as a US diplomat under Presidents Barack Obama and Joe Biden.

    Continue reading
  • IRS doesn't completely scrap facial recognition, just makes it optional
    But hey, new rules on deleting your selfies

    America's Internal Revenue Service has confirmed taxpayers will not be forced to use facial recognition to verify their identity. The agency also set out rules for which images will be deleted.

    Folks setting up an online IRS account will be given the choice of providing biometric data to an automated system, or speaking with a human agent in a video call, to authenticate. Those who are comfortable with facial recognition tech can upload a copy of their photo ID and then be authenticated by their selfie, and those who aren't can talk to someone to prove they are who they say they are. An online IRS account can be used to view tax documents and the status of payments among other things.

    "Taxpayers will have the option of verifying their identity during a live, virtual interview with agents; no biometric data – including facial recognition – will be required if taxpayers choose to authenticate their identity through a virtual interview," the IRS said in a statement on Monday.

    Continue reading
  • Sri Lanka to adopt India’s Aadhaar digital identity scheme
    Biometric IDs for all, cross-border interoperability not on the table

    Sri Lanka has decided to adopt a national digital identity framework based on biometric data and will ask India if it can implement that nation’s Aadhaar scheme.

    The island nation had previous indicated it would work with the Modular Open Source Identity Platform (MOSIP), an organisation based in India that offers tools governments can use to create and manage digital identities.

    But a list of Cabinet decisions published on Tuesday, Sri Lanka’s government announced its intention to ask India for a grant of its scheme, which has been widely interpreted as meaning India share Aadhaar technology.

    Continue reading

Biting the hand that feeds IT © 1998–2022