> daily_signal(2026_07_14)
LAPD's own inspector general counted 161 innocent drivers pulled over on false "stolen vehicle" hits from license-plate cameras in a single two-month window — the department has now disconnected its Flock cameras, and the Police Commission takes up the audit today.
PickBits Daily Signal · Tuesday, July 14, 2026
1. LAPD's own inspector general counted 161 innocent drivers pulled over on false "stolen vehicle" hits from license-plate cameras in a single two-month window — the department has now disconnected its Flock cameras, and the Police Commission takes up the audit today.
This is the cleanest public accounting yet of what an automated-surveillance false positive actually costs, and it comes from the department's own watchdog rather than a privacy group. The LAPD Office of the Inspector General audited the department's automated license-plate reader (ALPR) program over Aug 1 – Sep 30, 2025 and found 161 cases where officers confirmed the camera had matched the plate CORRECTLY — and the vehicle still was not stolen. That distinction is the whole story and it is the part most coverage flattens: this is not an AI accuracy problem. The optical character recognition worked. The system failed because the 'hot list' it checks against was stale — other jurisdictions had recovered vehicles and not removed them from the shared list in time — so the machine confidently and correctly matched a plate to a record that should no longer have existed. The downstream consequence of a database sync lag is a felony stop: an innocent driver, hands up, guns out. Do the arithmetic the audit invites: 161 false hits against 337 alerts that did recover a genuinely stolen vehicle means that of the 498 stolen-vehicle alerts officers acted on, roughly ONE IN THREE sent them after an innocent person. Set that against the scale — 210.5 million plate reads in those two months, and of 5,911 plates tracked, 4,575 produced no police action at all. Inspector General Matthew Barragan recommended the department stop installing new ALPR cameras and sign no new ALPR contracts pending public input, require annual audits, and state plainly that misusing ALPR data is a disciplinary offense. The department had not conducted a thorough ALPR audit since 2022; when a California Public Records Act request asked for audit records in 2025, LAPD said it had none. LAPD has since declined to renew its Flock Safety agreement and disconnected the 138 Flock pole cameras; the Axon contract expires at the end of July; and the Board of Police Commissioners takes the findings up today. The national frame: Flock alone runs on the order of 90,000–100,000 cameras reading roughly 20 billion plates a month, and the Supreme Court's Chatrie v. United States decision last month — the first major Fourth Amendment case in eight years — held that cell-phone location-history searches require a warrant, reasoning that courts should weigh what a surveillance technology is CAPABLE of, not merely what police did with it on the day. Legal scholars read that as pointed directly at ALPR. Flock's counter is that its cameras capture point-in-time images of vehicles in public view, not continuous tracking of a person.
Key fact: FIND OUT IF YOUR TOWN IS RUNNING ALPR, AND WHO IT SHARES WITH — this is a five-minute lookup most people never do. EFF's Atlas of Surveillance (https://atlasofsurveillance.org/) maps which US agencies have deployed ALPR, drones, face recognition and more, searchable by city and agency; the crowd-mapped DeFlock project (https://deflock.me/) plots individual ALPR camera locations and vendors. The sharing question matters more than the camera count: LA's false hits came from OTHER jurisdictions' stale records, so the risk to you is set by the worst data-hygiene practice among every agency your local department shares a hot list with. If you were pulled over on a plate-reader hit, an alert record exists — request it from the agency, and ask specifically which hot list the alert came from and when that entry was last validated. In Los Angeles, the Board of Police Commissioners meeting is public and its agendas are posted at https://www.lapdonline.org/police-commission/.
404media.co · nbclosangeles.com · futurism.com · therecord.media · lapdonline.org · primary source
2. Twenty-six current and former Meta workers sued in federal court on Monday, alleging Meta built its layoff list with a stack of internal AI systems fed by keystroke-and-screen monitoring — and that the scoring pushed out the people who had taken medical, family or pregnancy leave.
CONTINUING 2026-05-19 #3 — the day before Meta cut 8,000 people, we ran a story arguing that if the company said AI made the call, no federal law would require it to prove it. Twenty-six of those workers are now trying to force exactly that proof, and the vacuum we described is where this case will be fought. This is the case the entire 'AI in the workplace' debate has been waiting for, and the specific allegation is far sharper than 'a robot fired me.' The plaintiffs are not arguing that AI is unfair in the abstract. They are arguing a precise, testable mechanism of discrimination: Meta's scoring systems ranked employees on performance, productivity and output metrics — and an employee who was on legally protected leave produces fewer metrics BECAUSE THEY WERE LAWFULLY ABSENT. So the model, doing exactly what it was built to do, systematically scored them lower than peers who were at their desks. As the complaint puts it, the scoring 'not only failed to account for their protected leaves, but in effect penalized the employees for exercising their legal rights to these leaves.' That is the whole ballgame: a facially neutral metric becomes a discrimination engine the moment it treats a legally protected absence as an absence of merit. Every one of the 26 plaintiffs shares exactly one characteristic — all took, requested, or were approved for protected leave in the past 24 months. The named machinery is worth reading closely, because it is a catalogue of the ordinary enterprise AI stack rather than anything exotic: 'Metamate,' employee-trained 'second-brain' agents, keystroke- and activity-monitoring data, AI-TOKEN-USAGE DASHBOARDS, and algorithmically assisted performance ranking and calibration. Note that third one. If an internal dashboard measuring how much you use the company's AI tools is an input to whether you keep your job, then 'AI adoption' has quietly become a performance metric — and the person on parental leave is not burning tokens. The surveillance layer underneath is its own story: the plaintiffs describe a monitoring program capturing keystrokes, screen content, mouse activity, browser history, messages, emails, and voice, video and location data on company devices — announced, they say, through 'a low-visibility internal post — made by an engineer rather than a senior leader, in a secondary group rather than Meta's official employee-notice channel,' with no consent prompt on at least some teams and, initially, no way to opt out. And the data gathered that way was then used to BUILD the AI tools. There is a second, quieter lesson here for anyone who has ever clicked through an onboarding packet: because Meta requires a mutual arbitration agreement with a class-action waiver, these 26 people cannot bring a class action at all. They are pursuing their claims INDIVIDUALLY in arbitration, one at a time, and are in court only to ask for an injunction restoring their employment as of May 20. The arbitration clause did its job. Meta's denial is a single sentence and it is the sentence the entire case will turn on: 'Workforce management and organizational decisions were and are made by people, not AI.' Discovery will establish whether a human who rubber-stamps a machine-generated ranked list is a decision-maker or a signature.
Key fact: IF YOU TOOK PROTECTED LEAVE AND WERE THEN SELECTED IN A REDUCTION IN FORCE, THE LEGAL THEORY IN THIS COMPLAINT MAY BE YOURS TOO — AND THE CLOCK IS SHORTER THAN YOU THINK. The mechanism is now documented: output-based scoring penalizes the legally absent. Three concrete steps. (1) Read the EEOC's guidance on AI and the ADA (https://www.eeoc.gov/eeoc-disability-related-resources/artificial-intelligence-and-ada) — it establishes that an employer can violate the ADA when algorithmic tools screen out people with disabilities, including by failing to account for accommodations, and it applies to selection for termination, not just hiring. (2) Before you sign anything, DIG OUT YOUR ARBITRATION AGREEMENT. The 26 Meta plaintiffs are competent litigants with counsel and they still cannot bring a class action, because a mutual arbitration clause with a class-action waiver forces them into individual arbitration one by one. That clause is in most US tech offer letters and it is the single most consequential document in this story. Know which one you signed before you decide what leverage you have. (3) Severance agreements typically contain a release of claims with a signing deadline (often 21 or 45 days, with a 7-day revocation window under the OWBPA if you are 40+). Do not let it lapse uninvestigated if you were on FMLA, ADA or pregnancy leave in the scoring window — that is exactly the fact pattern here.
courthousenews.com · npr.org · foxbusiness.com · eeoc.gov · nbcnews.com · primary source
3. A Los Angeles Uber driver filed a class action in San Francisco alleging Uber profiles drivers with their own biometric and location data to compute the lowest fare each one will personally accept — what the complaint calls a "surveillance wage."
Personalized pricing has been the quiet endgame of consumer data collection for a decade; this complaint alleges Uber has already run it in reverse, on the supply side, against its own workforce. The allegation is not that Uber pays drivers badly. It is that Uber pays each driver a DIFFERENT, individually-computed amount for the same trip — an amount derived not from what the ride is worth or how well the driver performs, but from a machine-learned prediction of that specific person's reservation price: the least they, personally, will take. Lead plaintiff Edwin Carranza says the data he was required to hand over to work at all — precise location, trip history, acceptance and rejection patterns, cancellation history, driving times, app-use behavior, identity verification and biometric/facial-verification data — was fed back into a profile that then set his pay. He describes the resulting behavior with a specificity that will be easy to test in discovery, which is what makes this more than a vibes lawsuit: offers that got WORSE after he declined low-value rides; lower fares on trips heading in the direction of his home at the hour he usually stopped for the day; and surge notifications he says were frequently false, pulling him away from where he wanted to be. His framing is the one to keep: daily use of the app is a slot machine, with algorithms that 'distribute higher-value fares and lucrative bonuses unpredictably,' a 'sporadic reward system' that 'preys on hope' — a good day convinces you to work longer tomorrow, and then the payouts thin out. He calls the result a 'surveillance wage': compensation personalized on surveillance rather than performance. The legal architecture is interesting precisely because it sidesteps the swamp that has consumed gig-work litigation for years. Carranza is not relitigating employee-vs-contractor classification. He is suing on privacy and consumer-protection grounds — unfair competition, false advertising, invasion of privacy, intrusion upon seclusion, negligent misrepresentation, breach of the implied covenant of good faith and fair dealing, and unjust enrichment — and the core of it is CONSENT: he says he was never told, and never agreed, that his data would be used to build a model of his own price sensitivity. Uber's 2022 'Upfront Fares' rollout, sold as transparency (see your earnings before you accept), is recast in the complaint as the delivery mechanism for exactly this. Why an IT/policy audience should care beyond rideshare: the FTC's own surveillance-pricing study found that everything from your precise location to your mouse movements to what you abandoned in a cart is already being used by intermediaries to set individualized consumer prices. Uber's drivers are simply the population where the practice is most measurable, most consequential, and now most litigated. What Carranza asks for is the remedy that matters: an order forcing Uber to DISCLOSE what data and algorithmic factors set driver-facing trip amounts. Uber did not respond to a request for comment.
Key fact: IF YOU DRIVE FOR A PLATFORM, THE ONE MOVE THAT CHANGES YOUR LEVERAGE IS GETTING YOUR OWN DATA — most drivers never request it, and it is the only way to see what the model sees. Uber operates a privacy center (myprivacy.uber.com, reachable from the app under Privacy) where you can request a download of the data Uber holds on you, including trip and account data; California drivers have an additional statutory right under the CCPA/CPRA to request the specific pieces of personal information collected, the categories of sources, the business purpose, and the third parties it was shared with — and, importantly, that includes information used for automated decision-making and profiling. Pull it and keep it: the pattern Carranza describes (offers degrading after you decline low-value rides, worse fares heading home at your usual quitting hour, surge alerts that evaporate on arrival) is only provable against your own logged history. Whatever you conclude, keep your own independent record of offered vs. completed fares, because the platform's number is the thing in dispute.
courthousenews.com · ftc.gov · columbialawreview.org · hrw.org · primary source
4. The FDA just cleared an AI that screens for diabetic eye disease from a retinal photo a nurse can take in an ordinary doctor's office — no eye specialist in the loop — and it is on sale across the US now.
This is the constructive slot, and it earns it on the least glamorous axis there is: access. Diabetic retinopathy is one of the leading causes of preventable blindness, and the cruelty of it is that it is both silent and treatable — by the time vision changes are noticeable, damage is often permanent, while catching it early makes it manageable. The recommended defense is an annual dilated eye exam, and an enormous share of Americans with diabetes simply never get one. Not because they refuse. Because it means a separate appointment, with a separate specialist, often in a separate building, at a separate cost — and that friction is the disease's best ally. It falls hardest on exactly the people already least likely to see a specialist. What the FDA cleared on July 9 is a piece of software that attacks the friction rather than the disease. iHealthScreen — a small medical-device company based in Richmond Hill, New York, not a household-name lab — received 510(k) clearance for iPredict-DR, which analyzes color retinal fundus images captured on an iCare DRSplus camera and detects more-than-mild diabetic retinopathy in adults with diabetes who have not been diagnosed with it. The clearance is built on a clinical validation trial covering diagnostic performance, safety and usability. The load-bearing design decision is the one in the company's own description: the system is non-invasive and 'easy to use by any minimally skilled healthcare worker or nurse.' That sentence is the entire point. It means the screening can happen where the patient ALREADY IS — in the primary-care visit they were going to attend anyway, in a community clinic — instead of at a referral they were never going to make. Founder and CEO Alauddin Bhuiyan frames the mission as making 'AI-powered retinal screening accessible in primary care and community healthcare settings, enabling earlier detection, faster referral, and helping prevent avoidable vision loss.' Two honest caveats, because this section is where the hype usually goes. First, this is a SCREENING tool, not a treatment and not a substitute for ophthalmic care: a positive result is a referral, and the specialist is still the one who treats you. Second, the published coverage of the clearance does not disclose the sensitivity and specificity figures — those live in the FDA's 510(k) summary, and anyone deploying this should read them there rather than take a press release's word for it. But the shape of the thing is right, and it is the shape worth wanting: a narrow, regulated, human-supervised model, cleared by a regulator, deployed at the point of access, aimed squarely at a gap that costs people their eyesight. Set it against the other three stories on today's slate. Same underlying machinery — a model scoring a human being from data. The difference is a regulator in the loop, a disclosed purpose, and a person who gets something back.
Key fact: IF YOU OR SOMEONE YOU LOVE HAS DIABETES, THE ASK AT YOUR NEXT PRIMARY-CARE VISIT IS NOW A DIFFERENT ASK, AND IT TAKES ONE SENTENCE: 'Do you do the AI retinal screening here, or do I need a referral?' The single most common reason people with diabetes go blind from retinopathy is not that treatment failed — it is that the annual dilated eye exam never happened, because it meant another appointment with another specialist somewhere else. Autonomous AI retinal screening (iPredict-DR is the newest, but it is not the only FDA-cleared system on the US market) is specifically designed so the photo can be taken by a nurse or medical assistant during the visit you are already at, with results in the same sitting. Two things to hold onto: a screening result is NOT a diagnosis — a positive means you get referred to an eye specialist, and you should go — and a negative does not retire your eye care forever, it resets the clock on it. If your clinic does not offer it, ask whether they can refer you to a community health center that does, and ask your insurer whether the screening is covered (it commonly is, as it is a recommended annual preventive service for people with diabetes).
healio.com · ihealthscreen.org · accessdata.fda.gov · pharmashots.com · primary source