Facial Recognition

Senators Introduce Bill To Limit Facial Recognition Technology—but Does It Go Far Enough?

The legislation would require warrants for extended surveillance, but look at what it explicitly OKs.

|

A pair of senators are teaming up across the aisle to put limits on how federal law enforcement agencies use facial recognition tools and when they must seek a warrant.

This week Sens. Mike Lee (R–Utah) and Chris Coons (D–Md.) introduced the Facial Recognition Technology Warrant Act. If it becomes law, it would require federal officials to get a warrant if they're going to use facial recognition technology to attempt to track a specific person's public movements for more than 72 hours.

The bill does not prohibit the use of facial recognition technology to identify people or even to monitor events in real time. Indeed, it says the authorities can use facial recognition to identify people, even without a warrant, as long as "no subsequent attempt is made to track that individual's movement in real time or through the use of historical records after the individual has been identified."

In other words, the Facial Recognition Technology Warrant Act doesn't actually require warrants for the application of the technology, except for long-term surveillance of specific individuals.

Nevertheless, Coons and Lee put out a statement—joined by Fred Humphries, corporate vice president of U.S. government affairs at Microsoft—to praise what the bill would accomplish if passed. "Facial recognition technology can be a powerful tool for law enforcement officials," said Lee. "But it's [sic] very power also makes it ripe for abuse. That is why American citizens deserve protection from facial recognition abuse. This bill accomplishes that by requiring federal law enforcement agencies to obtain a warrant before conducting ongoing surveillance of a target."

Americans for Prosperity also declared its support of this bill, which it sees as more balanced than a full ban on government's use of facial recognition tech tools. Americans for Prosperity prefers regulations that keep authorities from abusing the tech. "We're standing behind this bill," said Senior Policy Analyst Billy Easley, "because we believe in the appropriate application of facial recognition technology and ensuring it is used for good rather than the mistreatment of Americans."

Other privacy activists are much less impressed. The bill doesn't stop the feds from accessing or using the hundreds of millions of face pictures they've already collected from drivers licenses and passports. In fact, it specifically gives such use a thumbs-up. At CNet, representatives of the American Civil Liberties Union and Fight for the Future express their misgivings:

"It has gaping loopholes that authorize the use of facial recognition for all kinds of abusive purposes without proper judicial oversight," said Evan Greer, Fight for the Future's deputy director. "It's good to see that Congress wants to address this issue, but this bill falls utterly short."

Read the bill for yourself here.

 

NEXT: Justin Amash to Trump: Let Bolton, Giuliani, and Mulvaney Testify

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. >>>get a warrant if they’re going to use facial recognition technology to attempt to track a specific person’s public movements for more than 72 hours

    behave for 71:59 at a time and crime it up in between. or can they track me for 71:59 then wait a second and start over?

  2. There are already trained neural nets out there that can identify a person based on gait and movement patterns. So facial recognition might not even be needed to identify someone on a surveillance camera.

    1. Nah, the gait analysis stuff only works from some relatively specific angles and in environments free of clutter. Without a relatively clean side-on shot (with about 20° of wiggle room) and roughly the same limit on declination (so high-angle traffic cameras are out) they can’t make very good identifications. They also don’t handle shadows very well. Not to mention you can fool them just by wearing platform soles, which is considerably less obvious than a mask (for the facial recognition stuff).

      I mean, current facial recognition tech isn’t really that great either, but the easiest methods for foiling it are at least apparent at a glance to regular cops, so some kind of ban on face coverings or weird glitchy camo could be enforceable to a certain extent. Personally I think that the best limitations on facial recognition tech would be to ban its use in obtaining warrants, since it’s got a fairly high false positive rate and the cops already play fast and loose with the evidentiary rules for warrants.

  3. This seems like rather weak sauce. Especially considering the practice of parallel construction.

    1. Agreed: Until we find some way to crack down on parallel construction, all the courts are doing is restricting what they can admit to having done. And non-existent tipsters will continue to tell them exactly where to look for admissible evidence, once the illegal means have told them what the fictional tipster should say.

  4. Wait until the AI focuses on women’s chests and not their faces.
    You just wait. You’ll see….

    1. First question is how can the AI tell it is a woman? Even DNA cannot tell that now. It is fully subjective, and internal the the carbon based life form involved.

  5. “Senators Introduce Bill To Limit Facial Recognition Technology—but Does It Go Far Enough?”

    @Reason seems to support banning facial recognition technology BUT NOT facial recognition – lets say by eye-sight.

    It implies a violation of privacy but has yet to explicitly explain how that privacy is violated.

  6. “It has gaping loopholes that authorize the use of facial recognition for all kinds of abusive purposes without proper judicial oversight,” said Evan Greer, Fight for the Future’s deputy director. “It’s good to see that Congress wants to address this issue, but this bill falls utterly short.”

    Sounds typical, preventing the government from doing illegal things by making the things it wants to do legal.

  7. ““Senators Introduce Bill To Limit Facial Recognition Technology—but Does It Go Far Enough?””

    So @Reason does support banning things after all ?

  8. >look at what it explicitly OKs

    “Look at”? Wow, reason is the wizard of context.

    This comment not approved by Silicon Valley brain slugs.

  9. Disclosure: I am not a lawyer.

    After skimming the first 2/3 or so of the bill linked at the end, my sense is that this is misdirected and too particular.

    The problem, if there is one, concerns surveillance as such, rather than surveillance using this particular method or that. It would make sense that the same rules should apply to any surveillance, by whatever technique. I have no doubt that there are generally accepted common law, and maybe statutory, rules for surveillance that could be codified if necessary, and tweaked to ensure that they cover recent and anticipated new techniques.

    It would be better law to target, and describe usage restrictions upon, all government surveillance. There is no difference in principle between using AI, however it may be defined at a point in time, to search stored or live photographic or video data, from using live (presumably) human intelligence to search video tapes or mug books of hard copy photographs. The real issues involve the scope of the dataset being searched, its origin (e. g., convicts, arrestees, those present on a street or other public place, or state ID holders), and only secondarily the method of search. A lot has been written about AI limitations resulting from algorithmic and training data sets, and most of it is pretty much rubbish absent context that includes things like accuracy of human searchers and well documented unreliability of witness reports. My suspicion is that the best AI facial recognition now performs better than humans on general sets of pictures or videos and exhibits less of what generally might be described as race (or sex) bias. Any research on this question that may exist it is notable by its lack of mention in general press reports.

    This particular piece of legislation has the appearance of a quickly conceived band-aid likely to be inadequate and fail before too long. It should be binned and ultimately replaced by well thought out, and broader scoped, legislation.

  10. It’s good to see that Congress wants to address this issue, but this bill falls utterly short.

    Unfortunately – there doesn’t seem to be any way for people to actually force changes or alterations in this sort of legislation. There’s no such as competitive elections for representation. Even if there was, our susceptibility to be manipulated en masse via partisanship and money makes it impossible to even elect a small handful of critters who might be able to work from the inside of a legislature.

    1. Like many things that “everyone knows,” the claim that there are no competitive elections is false. Nearly all Congressional districts are contested by at least the two major parties, and many are contested by additional parties or independent candidates. Moreover, most states have primary elections in which voters choose a party’s candidate from two or more aspirants. While it often is true that the major party candidates differ little, that is no more than an accurate reflection of preferences in the electorate. A considerable number of districts, however, are genuinely competitive, particularly when the advantage of incumbency – the known devil – is absent because of retirement or redistricting.

      The fact that there is not great pressure for change is that the great majority of the population regards the status quo as “good enough,” so much so, indeed, that many do not bother to vote. Those who study the national lead press or follow CSpan are a small, though vocal, minority.

      The major part of the voting eligible population is ignorant of the way the many governments they are subject to work, largely ignorant of the issues in any meaningful sense, and tend to be skeptical or even dismissive of candidates’ promises. They are motivated by no more than three or four issues, many of them more or less local in character. As often as not, or more, the most important motivating issue still is the name of the political party they learned as small children from their parents, and either kept or rejected in adolescence.

      1. Like many things that “everyone knows,” the claim that there are no competitive elections is false…A considerable number of districts, however, are genuinely competitive,

        There are seven different groups who actually assess the probability of a competitive election in every district. Usually as part of some consulting work they do for big donors and/or parties. In 2018, 294 of the House districts were deemed to have no chance of competition (safe or solid) by all seven groups. 24 more were deemed safe/solid by 5+ of those 7 with only 1 or 2 deeming it ‘likely’ (ie not ‘competitive’ at all absent goat porn pictures). That’s 73% of the population that does not in any remote sense at all have any choice about who their critter will be.

        Only 46 seats had even 1 of those entities deem the race a ‘tossup’ (the only level of competition where an individuals vote might actually matter). 16 of those were open seats (critter retiring) so the competition there does not affect an actual incumbent. That leaves 30 districts – or 7% of the US population – who had a predictably significant chance to change their incumbent in a midyear election (where the election is really about that level critter rather than the Prez race). In 2018, 22 of those 30 lost and 8 won.

        I don’t call 7% even remotely ‘considerable’.

        1. Moreover, even a nominally “competitive” election doesn’t help any on topics the major parties actually agree on. For that you need to get into the primaries, and the primaries are hardly ever competitive if there’s an incumbent of that party in office already.

          The truth is that we have a self-perpetuating political class in this country, and they’re very good at restricting holding elective office to their own members. That’s one of the reasons Trump is facing such hostility: He didn’t come from within their numbers.

    2. Good grief, your rant is like reading someone explain chemistry in terms of the four elements of alchemy. What you write isn’t even wrong, it’s just nonsense.

      How about you simply pick up some books on political science and figure out how things actually work?

  11. Anything in the public arena has no reasonable expectation of privacy. That’s the way I understand 4th protections on privacy. So, police can follow someone in their car for days, weeks, or even years to conduct surveillance on a person of interest, but now that they want a computer to do the same thing, it’s wrong? I vehemently disagree with this law.

    1. So, police can follow someone in their car for days, weeks, or even years to conduct surveillance on a person of interest, but now that they want a computer to do the same thing, it’s wrong?

      You can follow anybody you want in public.

      For government, any action needs to fall within one of the enumerated powers. Following citizens around absent just cause isn’t among those powers.

  12. I have received $21746 in my last 4 weeks by working online from home just in my part time. I have joined this job 2 months ago and in my first month i have received $20159 easily without having any online working experience. Everybody can get this job today and start making real money online by follow instructions on this website.

    http://buzzjobs.com.nu/

Please to post comments