{"meta":{"exportedAt":"2026-11-07T02:11:06.448Z","formatVersion":2,"selection":{"q":"work","system":"","harm":"","context":"","country":"","role":"","relation":"","evidence":"","year":"2021","response":"","severity":"","verification":"","view":"incidents","sort":"added"},"totalIncidents":5,"coverage":{"cases":5,"countries":2,"languages":1,"unknownLocation":0,"locationPending":0,"unknownLanguage":0,"unknownDate":0,"lawsuits":1,"regulatory":0,"minors":0,"coreRelations":0,"contextualRelations":5,"mixedRelations":0,"unknownRelations":0,"relationPending":0,"relationUnknown":0},"countingNote":"Distinct public cases in this selection. People counts apply within individual cases only; cross-case person overlap has not been resolved. No population incidence estimate.","affectedCountNote":"Interpret person counts with affectedCountStatus and the reported effects. Unquantified zeros are placeholders, not a measured zero.","source":"AI incidents","publisher":"NOPE","url":"https://nope.net/incidents","license":"CC BY 4.0"},"incidents":[{"id":"2021-massachusetts-cvs-job-applicant-alleges-hirevue-ai-video-interview-was-unlawful-lie-detector-test-court-lets-claim-proceed","caseFacts":{"claims":[{"id":"c1","status":"reported","evidence":[{"locator":"'To   screen    applicants,    CVS   administers      a    video-interview' ... 'platform    called    Affectiva,    which    analyzes     candidates’     facial' ... 'expressions, eye contact, voice intonation, and inflection using' ... 'artificial intelligence.'","relation":"supports","source_id":"s1"},{"locator":"'video of their facial expressions, eye contact, voice intonation, and inflection are analyzed using artificial intelligence.'","relation":"supports","source_id":"s2"}],"assertion":"The complaint alleges that CVS screens applicants with a HireVue video interview, that HireVue uploads the recordings to Affectiva, which analyses facial expressions, eye contact, voice intonation and inflection using artificial intelligence, and that HireVue gives CVS a numerical score for each applicant.","causal_attribution":"The applicant's allegations as summarised by the court and a law firm; not tested in court."},{"id":"c2","status":"reported","evidence":[{"locator":"'Around January 2021, Baker applied for a supply chain position in Massachusetts with CVS.' ... 'did not include the notice of rights mandated' ... 'During    Baker’s     application     process,       CVS   administered    a' ... 'he would not have participated in the'","relation":"supports","source_id":"s1"}],"assertion":"Around January 2021 CVS administered a HireVue interview to the applicant for a Massachusetts supply chain position; the online application lacked the statutory notice, and he says he did not know the interview was a lie detector test and would not have taken part had he known.","causal_attribution":"The applicant's allegations accepted as true for the motion."},{"id":"c3","status":"reported","evidence":[{"locator":"'Baker did not get hired and then sued'","relation":"supports","source_id":"s2"}],"assertion":"According to a law firm commentary, the applicant was not hired.","causal_attribution":"A law firm's summary; the court's order does not state the hiring outcome, and no source connects the decision to an AI score."},{"id":"c4","status":"documented","evidence":[{"locator":"'After' ... 'hearing, the Court DENIES CVS’s motions to dismiss Count III'","relation":"supports","source_id":"s1"},{"locator":"'Thus, CVS’s Motion to Dismiss was denied'","relation":"supports","source_id":"s2"}],"assertion":"On 16 February 2024 the federal court denied CVS's motions to dismiss the notice claim, holding that the statute lets aggrieved applicants sue and that the applicant had pleaded a concrete informational injury.","causal_attribution":"Established by the court's own order."},{"id":"c5","status":"reported","evidence":[{"locator":"'CVS Health Corp. has reached an individualized settlement with a job applicant who alleged the pharmacy chain broke Massachusetts law by subjecting him to a lie detector test through AI-aided video-interview technology.'","relation":"supports","source_id":"s3"}],"assertion":"In July 2024 CVS reached an individual settlement with the applicant.","causal_attribution":"Reported by Bloomberg Law; the terms were not reported and the settlement is not an admission."}],"effects":[{"label":"Applicant says he took part in an AI-analysed interview he alleges was a lie detector test, without the notice the law requires","claim_id":"c2","direction":"negative"},{"label":"Applicant was not hired, according to a law firm commentary","claim_id":"c3","direction":"negative"}],"sources":[{"id":"s1","url":"https://www.govinfo.gov/content/pkg/USCOURTS-mad-1_23-cv-11483/pdf/USCOURTS-mad-1_23-cv-11483-0.pdf","kind":"court_record","access":"read","language":"en","translation_note":"Read in English on 2026-11-02: the 16 Feb 2024 memorandum and order (PDF, GovInfo, HTTP 200). Its factual background restates the complaint's allegations, which the court accepts as true for the motion only.","independence_group":"d-mass-baker-v-cvs"},{"id":"s2","url":"https://www.afslaw.com/perspectives/ai-law-blog/using-new-technology-vet-job-applicants-can-violate-the-massachusetts-lie","kind":"news_report","access":"read","language":"en","translation_note":"Read in English on 2026-11-02 (HTTP 200). A law firm's commentary on the ruling; it summarises the complaint and the ruling and adds that the applicant was not hired.","independence_group":"arentfox-schiff-2024-05-20"},{"id":"s3","url":"https://news.bloomberglaw.com/litigation/cvs-applicant-agrees-to-settle-ai-aided-lie-detector-test-suit","kind":"news_report","access":"read","language":"en","translation_note":"Read in English on 2026-11-02 (HTTP 200): only the opening paragraphs are visible without a subscription; they report the individual settlement.","independence_group":"bloomberg-law-2024-07-18"}],"version":1,"ai_roles":["institutional_use"],"contexts":["work"],"unknowns":["Whether Affectiva's AI analysis actually ran on the applicant's January 2021 interview, and what score or report HireVue gave CVS about him.","Whether any AI score affected CVS's decision not to hire him.","The terms of the July 2024 settlement.","CVS's and HireVue's own account of the interview: neither inspected source reports their response to the allegations about the AI analysis."],"geography":{"basis":"The court says he applied for a supply chain position in Massachusetts with CVS. The court is the US District Court for the District of Massachusetts. No residence is used.","court_countries":["US"],"event_countries":["US"],"affected_person_countries":[]},"publication":{"basis":"Published because a federal court order records the applicant's allegation that CVS gave him a HireVue interview whose recording was analysed with AI to score applicants, and denied CVS's motion to dismiss his claim of a concrete injury from taking it without the required notice; the case later settled. The AI analysis remains an untested allegation.","reviewed_on":"2026-11-02"},"ai_involvement":{"basis":"The complaint, as summarised by the court, alleges that CVS administered a HireVue video interview to the applicant, that HireVue sends interview recordings to Affectiva, which analyses facial expressions, eye contact, voice intonation and inflection using artificial intelligence, and that HireVue gives CVS a numerical score for each applicant, including assessments of honesty and integrity. The harm alleged is taking part in that assessment without the statutory notice. These are allegations accepted as true only for the motion to dismiss; no source establishes that the AI analysis ran on his interview or what it concluded.","status":"reported"},"person_relations":["made_claim_about"]},"name":"Massachusetts job applicant alleges CVS's AI-analysed HireVue video interview was an unlawful lie detector test; court let his claim proceed, then he settled","summary":"A man who applied for a supply chain job with CVS in Massachusetts around January 2021 says CVS put him through a HireVue video interview with questions about integrity, and alleges that HireVue sent the recording to a platform called Affectiva, which analysed his facial expressions, eye contact, voice intonation and inflection with artificial intelligence and fed a score to CVS. He sued CVS in a proposed class action under the Massachusetts Lie Detector Statute, alleging the interview was a lie detector test and that the online application lacked the notice of rights the law requires. He says he did not know the interview was a lie detector test and would not have taken part had he known. In February 2024 a federal judge in Massachusetts denied CVS's motion to dismiss the notice claim, finding a concrete informational injury; CVS had not challenged the other claims at that stage. A law firm commentary says he was not hired. In July 2024 he reached an individual settlement with CVS, whose terms were not reported. The allegations about the AI analysis were not tested in court.","incidentDate":"2021-01-01","incidentKind":"single_event","incidentDatePrecision":"month","exposurePattern":"single_interaction","reportedDate":"2024-02-16","aiSystem":"HireVue video-interview platform that, the applicant alleges, sent interview recordings to Affectiva for AI analysis of facial expressions, eye contact and voice, and gave CVS an employability or competency score for each applicant","aiProduct":"HireVue (reported)","severity":"low","verificationStatus":"credible","harmCategories":[],"harmOutcomes":["professional_harm"],"harmOutcomeSummary":"The applicant alleges CVS subjected him, without the notice Massachusetts law requires, to a HireVue video interview analysed by AI that he says was a lie detector test; a law firm commentary says he was not hired. The court found a concrete informational injury pleaded.","frameworkFacets":[],"causationStatus":"alleged","participantUsersAffectedMin":1,"otherPeopleHarmedMin":0,"affectedCountStatus":"exact","affectedCountEvidence":"One person: the applicant who took the HireVue interview and sued. The proposed class of Massachusetts applicants was never certified and is not counted.","victimAgeRange":"adult","jurisdiction":"US","platformType":"other","outcomeType":"lawsuit_settled","outcomeStatus":"resolved","primarySourceUrl":"https://www.govinfo.gov/content/pkg/USCOURTS-mad-1_23-cv-11483/pdf/USCOURTS-mad-1_23-cv-11483-0.pdf","primarySourceLabel":"US District Court, D. Mass., Baker v. CVS Health Corp., No. 23-11483, memorandum and order of 16 Feb 2024 (GovInfo)","firstPublishedAt":"2026-11-02T03:08:06.332184+00:00","updatedAt":"2026-11-02T03:08:06.332184+00:00","scopeVersion":"facts-v3","tags":["work","hiring","ai-video-interview","lie-detector-statute","lawsuit"]},{"id":"2021-nevada-unemployment-claimant-says-id-me-face-check-rejected-him-and-locked-him-out-of-benefits","caseFacts":{"claims":[{"id":"c1","status":"reported","evidence":[{"locator":"which uses a combination of biometric information and official documents to confirm that applicants are who they claim to be.; ID.me CEO Blake Hall told Motherboard that the company’s facial recognition technology does one-to-one matching—comparing one face against a picture of that same face (from a driver’s license, say); At least 21 states use ID.me to detect fraudulent benefit claims.","relation":"supports","source_id":"s1"}],"assertion":"ID.me verifies the identity of people claiming unemployment benefits in at least 21 US states, using biometric information and official documents, including a facial recognition step that compares one face with a picture of that same face, such as from a driving licence.","causal_attribution":"Motherboard's description and ID.me's chief executive's account of the system."},{"id":"c2","status":"reported","evidence":[{"locator":"a gig economy worker in Las Vegas who was suddenly cut off from his unemployment benefits in late March after ID.me failed to identify him.; it rejected it, didn’t give us a reason, just rejected it. It rejected it three times, and then it locked me out of the system.","relation":"supports","source_id":"s1"}],"assertion":"A gig economy worker in Las Vegas says he was suddenly cut off from his unemployment benefits in late March 2021 after ID.me failed to identify him; he says the facial recognition test rejected him three times without giving a reason and then locked him out of the system.","causal_attribution":"The claimant's account to Motherboard; the year comes from the article's date of 18 June 2021."},{"id":"c3","status":"reported","evidence":[{"locator":"He tried contacting the state of Nevada for help, but the employees there directed him back to ID.me.; until he tweeted a scathing criticism of the company, which then reached out and—after several more frustrating days—verified","relation":"supports","source_id":"s1"}],"assertion":"The claimant says ID.me's chat support did not resolve the problem despite many attempts and the state of Nevada directed him back to ID.me; after several weeks he posted a public criticism, and the company contacted him and, after several more days, verified his identity.","causal_attribution":"The claimant's account as reported by Motherboard."},{"id":"c4","status":"reported","evidence":[{"locator":"went for three weeks without receiving his benefit. “I couldn’t pay bills,”","relation":"supports","source_id":"s1"}],"assertion":"The claimant says he went three weeks without his benefit and could not pay bills.","causal_attribution":"The claimant's own statement to Motherboard."},{"id":"c5","status":"reported","evidence":[{"locator":"“The algorithms used for Face Match operate ~99.9% efficacy,” Hall wrote in an email to Motherboard.; Hall said that facial recognition failures are not a problem with the technology but with the people using it to verify their identity.; “We are unaware of eligible individuals who have not been able to verify their identity with ID.me,”","relation":"supports","source_id":"s1"},{"locator":"ID.me chief Blake Hall blamed the problems on users rather than the technology. Face Match algorithms have \"99.9% efficacy,","relation":"supports","source_id":"s2"}],"assertion":"ID.me's chief executive told Motherboard that its Face Match algorithms operate at about 99.9% efficacy, that face-match failures are a problem with the people using the system rather than the technology, and that the company was unaware of eligible individuals unable to verify their identity.","causal_attribution":"ID.me's response, kept as counterevidence; Engadget relays the same statement."},{"id":"c6","status":"reported","evidence":[{"locator":"ID.me unsurprisingly took issue with the Motherboard piece in a response provided to Engadget. Hall called the fraud claims \"inaccurate and defamatory,\" and stressed that the company suck to National Institute of Standards and Technology standards for identification. He reiterated the claims made about the nature of the facial recognition system and the video chat backup","relation":"supports","source_id":"s2"}],"assertion":"In a response to Engadget, ID.me called Motherboard's fraud claims inaccurate and defamatory and repeated its account of the facial recognition system and its video chat backup.","causal_attribution":"ID.me's statement to Engadget."}],"effects":[{"label":"claimant says the face check rejected and locked him out, stopping his benefits for three weeks","claim_id":"c2","direction":"negative"}],"sources":[{"id":"s1","url":"https://www.vice.com/en/article/facial-recognition-failures-are-locking-people-out-of-unemployment-systems/","kind":"news_report","access":"read","language":"en","translation_note":"Read in English on 2026-10-30; no translation.","independence_group":"motherboard-2021-06-18"},{"id":"s2","url":"https://www.engadget.com/facial-recognition-failures-id-me-unemployment-benefits-172654494.html","kind":"news_report","access":"read","language":"en","translation_note":"Read in English on 2026-10-30; no translation.","independence_group":"engadget-2021-06-19"}],"version":1,"ai_roles":["institutional_use"],"contexts":["public_services","finance"],"unknowns":["The exact dates of the lockout and of the verification; Motherboard gives only late March and three weeks without the benefit.","Why the face match failed in this case: no analysis of the attempts is reported, and ID.me attributes failures in general to users' photos.","Whether the missed benefit weeks were later paid."],"geography":{"basis":"The benefits were Nevada unemployment benefits: the claimant says he sought help from the state of Nevada, which sent him back to ID.me. Motherboard describes him as a gig economy worker in Las Vegas.","court_countries":[],"event_countries":["US"],"affected_person_countries":["US"]},"publication":{"basis":"Published as a state unemployment system's facial recognition identity check rejecting a described claimant three times and locking him out, which he says stopped his benefit for three weeks until the company verified him. His account is attributed; ID.me's explanation that failures come from users is recorded.","reviewed_on":"2026-10-30"},"ai_involvement":{"basis":"Motherboard and ID.me's chief executive describe ID.me Face Match as a facial recognition step that compares a claimant's face with the photo on an identity document. The claimant says that step rejected his face three times and locked him out of the system, which cut off his unemployment benefit, and that ID.me later verified him as the same person after his public complaint. The rejection of a person the company later verified is the system's own failed match. ID.me's chief executive says such failures come from users' photos; the cause in this case is not established.","status":"reported"},"person_relations":["made_decision_about"]},"name":"Nevada unemployment claimant says ID.me face check rejected him three times and locked him out, leaving him three weeks without benefits","summary":"ID.me, an identity-verification company used by at least 21 US states to check people claiming unemployment benefits, asks claimants to pass a facial recognition step that compares their face with a photo of the same person, such as a driving licence. In June 2021 Motherboard (VICE) reported that a gig economy worker in Las Vegas was suddenly cut off from his unemployment benefits in late March after ID.me failed to identify him. He said that when he held his phone in the instructed position the system rejected him three times without giving a reason and then locked him out. He said he tried ID.me's chat support many times and the state of Nevada sent him back to ID.me; after several weeks he posted a public criticism on Twitter, and the company contacted him and, after several more days, verified his identity. He said he went three weeks without his benefit and could not pay bills. ID.me's chief executive told Motherboard that its Face Match algorithms operate at about 99.9% efficacy, that failures are a problem with how people use the system rather than the technology, and that the company was unaware of eligible people unable to verify their identity. Motherboard said other claimants in several states described similar lockouts; those accounts are not counted here.","incidentDate":"2021-03-01","incidentKind":"single_event","incidentDatePrecision":"month","exposurePattern":"repeated_interactions","reportedDate":"2021-06-18","aiSystem":"ID.me Face Match, a one-to-one facial recognition step in ID.me's identity verification for state unemployment benefit claimants, as described by Motherboard and ID.me's chief executive","aiProduct":"ID.me Face Match","aiCompany":"ID.me","severity":"low","verificationStatus":"credible","harmCategories":[],"harmOutcomes":["financial_loss","psychological_distress"],"harmOutcomeSummary":"The claimant says ID.me's face check rejected him three times and locked him out, cutting off his unemployment benefit; he says he went three weeks without it, could not pay bills, and spent weeks trying to reach ID.me and the state before the company verified him. He describes the experience as ridiculous.","frameworkFacets":[],"causationStatus":"alleged","participantUsersAffectedMin":1,"otherPeopleHarmedMin":0,"affectedCountStatus":"exact","affectedCountEvidence":"One claimant describes being rejected by the face check and losing benefits. Motherboard says other claimants in several states complained of similar problems but gives no count tied to the face check, so they are not counted.","victimAgeRange":"adult","jurisdiction":"US","platformType":"other","outcomeType":"media_coverage","outcomeStatus":"resolved","primarySourceUrl":"https://www.vice.com/en/article/facial-recognition-failures-are-locking-people-out-of-unemployment-systems/","primarySourceLabel":"Motherboard (VICE), Facial Recognition Failures Are Locking People Out of Unemployment Systems, 18 June 2021","firstPublishedAt":"2026-10-30T03:05:29.391311+00:00","updatedAt":"2026-10-30T03:05:29.391311+00:00","scopeVersion":"facts-v3","tags":["id-me","facial-recognition","unemployment-benefits","identity-verification","public-services","nevada"]},{"id":"2021-amazon-delivery-drivers-say-netradyne-ai-van-cameras-flagged-driving-events-they-did-not-commit-lowering-bonus-scores","caseFacts":{"claims":[{"id":"c1","status":"reported","evidence":[{"locator":"'In February, Amazon announced that it would install cameras made by the AI-tech startup Netradyne in its Amazon-branded delivery vans'; 'Each time the camera registers an event, footage is uploaded into a system, recorded, and affects a score drivers receive at the end of the week for safe driving.'","relation":"supports","source_id":"s1"}],"assertion":"Amazon installed AI-powered Netradyne cameras in its delivery vans during 2021; the cameras detect driving events, upload the footage and feed a weekly safe-driving score.","causal_attribution":"Describes the system and its role in the score."},{"id":"c2","status":"reported","evidence":[{"locator":"'the camera in his van started to incorrectly penalize him whenever cars cut him off'; 'the Netradyne camera in his van has also detected “events” that didn’t actually happen, but that impacted his performance score at Amazon'","relation":"supports","source_id":"s1"}],"assertion":"Drivers said the cameras registered events they did not commit or that were not unsafe: a driver in Los Angeles said it incorrectly penalised him when cars cut him off, and a former driver in Texas said it detected events that did not happen and that they affected his performance score.","causal_attribution":"The drivers' accounts attribute the penalties to the camera's detections; Amazon and Netradyne did not address these individual events."},{"id":"c3","status":"reported","evidence":[{"locator":"'A few times, we’ve been in the country on a dirt road, where there’s no stop sign, but the camera flags a stop sign.'; 'If I look into my mirrors to make sure I am safe to change lanes, it dings me for distraction because my face is turned to look into my mirror.'","relation":"supports","source_id":"s1"}],"assertion":"Other drivers described false stop-sign detections, including a stop sign flagged on a dirt road with no sign, and being flagged for distraction when checking mirrors.","causal_attribution":"Drivers' accounts of the cameras' detection errors."},{"id":"c4","status":"reported","evidence":[{"locator":"'For many Amazon drivers, these performance scores determine whether they receive weekly bonuses, prizes, and extra pay.'; 'Before I would be able to win prizes and stuff, as soon as cameras came along, it went downhill,'; 'which receive bonuses by earning “fantastic” scores on a weekly scorecard'","relation":"supports","source_id":"s1"}],"assertion":"According to drivers and an Amazon scorecard document obtained by Motherboard, the weekly scores determine bonuses, prizes and extra pay, and the former Texas driver said winning prizes went downhill once the cameras came.","causal_attribution":"Connects the detections to pay; the amount lost by any individual driver is not reported."},{"id":"c5","status":"reported","evidence":[{"locator":"'but they don’t respond. There’s no room for discussion around the possibility that maybe the camera’s data isn’t clean.'; 'made the appeal process futile.'","relation":"supports","source_id":"s1"}],"assertion":"The Los Angeles driver said requests through his delivery company to see the photos flagged as events went unanswered, and a delivery-company owner said appeals were futile.","causal_attribution":"The drivers' and owner's account of the appeal route."},{"id":"c6","status":"reported","evidence":[{"locator":"'a team of Amazon employees manually reviews all events that are appealed to ensure that erroneous events do not impact drivers or Amazon delivery companies'; 'This technology provides drivers real-time alerts to help them stay safe when they are on the road.'; 'accidents had decreased by 48 percent'","relation":"supports","source_id":"s1"}],"assertion":"Amazon said the cameras give real-time safety alerts, that accidents had fallen 48 percent since installation, and that its employees manually review all appealed events so that erroneous events do not affect drivers.","causal_attribution":"Amazon's response; its safety figures are its own claims."}],"effects":[{"label":"Drivers say false camera events lowered their weekly scores","claim_id":"c2","direction":"negative"},{"label":"Scores decide bonuses and prizes","claim_id":"c4","direction":"negative"}],"sources":[{"id":"s1","url":"https://www.vice.com/en/article/amazons-ai-cameras-are-punishing-drivers-for-mistakes-they-didnt-make/","kind":"news_report","access":"read","language":"en","translation_note":"Read in English on 2026-10-27 (HTTP 200, full article body).","independence_group":"motherboard-netradyne-2021-09-20"}],"version":1,"ai_roles":["institutional_use"],"contexts":["work","finance"],"unknowns":["How many events recorded against these drivers were wrong, and whether any was later overturned on appeal.","How much bonus or prize value any individual driver lost.","Whether Amazon or Netradyne changed the detection thresholds after these reports."],"geography":{"basis":"The drivers worked from Amazon delivery stations in Los Angeles, Sugar Land (Texas), Mobile (Alabama), Oklahoma and Kentucky, and the vans and routes are in the United States.","court_countries":[],"event_countries":["US"],"affected_person_countries":["US"]},"publication":{"basis":"Published because the drivers describe specific detection errors by the AI cameras (cars cutting in scored as unsafe following, mirror checks scored as distraction, a stop sign flagged where none existed) and say those detections lowered the weekly scores that decide their bonuses and prizes. The accounts are attributed, Amazon's response and appeal process are recorded, and the amount lost is left unknown.","reviewed_on":"2026-10-27"},"ai_involvement":{"basis":"Motherboard describes the Netradyne cameras as AI-powered devices that detect driving events (following distance, stop-sign and signal violations, distracted driving) and upload them to a score. The drivers say the cameras' own detections were wrong: a car cutting in was recorded as unsafe following distance, a glance at a mirror as distraction, and a stop sign was flagged on a road that had none. Those detections entered the weekly scores that, according to the drivers and Amazon's internal scorecard document, decide bonuses and prizes.","status":"reported"},"person_relations":["made_claim_about"]},"name":"Amazon delivery drivers say Netradyne AI cameras in their vans flagged driving 'events' they did not commit, lowering the weekly scores that set bonuses","summary":"In September 2021 Motherboard reported accounts from six Amazon delivery drivers in five US states and the owner of an Amazon delivery company, who said the AI-powered Netradyne cameras Amazon had installed in its delivery vans from early 2021 regularly registered driving 'events' that were beyond their control or were not unsafe driving: being cut off by another car, glancing at a side mirror, stopping past a stop sign to see around a bush, or a stop sign flagged on a dirt road with no sign. Each event fed a weekly safe-driving score that, according to the drivers and an internal Amazon document Motherboard obtained, helped decide drivers' bonuses and prizes and delivery companies' bonuses. A driver in Los Angeles said the camera incorrectly penalised him and that requests to see the photos flagged as events went unanswered; a former driver in Texas said events that did not happen affected his score, which decided prizes from his delivery company. Amazon said the technology gives drivers real-time safety alerts, that accidents and violations had fallen since it was installed, and that its staff manually review appealed events so that erroneous ones do not affect drivers. Netradyne did not respond to Motherboard.","incidentDate":"2021-01-01","incidentKind":"ongoing_experience","incidentDatePrecision":"year","exposurePattern":"repeated_interactions","reportedDate":"2021-09-20","aiSystem":"Netradyne AI-powered four-lens cameras installed in Amazon-branded delivery vans, which detect driving 'events' such as following distance, stop-sign and signal violations and distracted driving, upload the footage and feed Amazon's weekly driver and delivery-company scores","aiProduct":"Netradyne cameras","severity":"low","verificationStatus":"credible","harmCategories":[],"harmOutcomes":["professional_harm","financial_loss"],"harmOutcomeSummary":"Drivers told Motherboard that the AI cameras registered events they did not commit or that were not unsafe, lowering the weekly scores that decided their bonuses and prizes; one said winning prizes 'went downhill' once the cameras came, and another described the penalties as upsetting.","frameworkFacets":[],"causationStatus":"alleged","participantUsersAffectedMin":0,"otherPeopleHarmedMin":2,"affectedCountStatus":"partial","affectedCountEvidence":"Two drivers are counted: the Los Angeles driver, who said the camera incorrectly penalised him in the data used for his weekly evaluation and bonus, and the former Sugar Land driver, who said events that did not happen affected his score and his prizes. Motherboard spoke to six drivers and a delivery-company owner who described false or unfair events more generally; the others are not counted because a personal score or pay consequence is not reported for each. The fleet size and driver numbers in the article are not counts of harmed people. The drivers are not counted as direct AI users.","victimAgeRange":"adult","jurisdiction":"US","platformType":"other","outcomeStatus":"unknown","primarySourceUrl":"https://www.vice.com/en/article/amazons-ai-cameras-are-punishing-drivers-for-mistakes-they-didnt-make/","primarySourceLabel":"Motherboard (VICE), 20 September 2021: Amazon's AI cameras are punishing drivers for mistakes they didn't make","firstPublishedAt":"2026-10-27T03:09:21.159219+00:00","updatedAt":"2026-10-27T03:09:21.159219+00:00","scopeVersion":"facts-v3","tags":["work","delivery-drivers","workplace-monitoring","in-vehicle-camera","driver-scores","pay","institutional-use"]},{"id":"2021-telangana-samagra-vedika-algorithm-wrongly-tagged-poor-families-as-car-owners-food-security-cards-denied","caseFacts":{"claims":[{"id":"c1","status":"reported","evidence":[{"locator":"'started arming the officials with Samagra Vedika, an algorithmic system that consolidates citizens'; 'the officials were mandated to check whether the algorithmic system approved the eligibility of the applicant before making their own decision'; 'They could either go with the algorithm'","relation":"supports","source_id":"s1"}],"assertion":"Since 2016 Telangana has used Samagra Vedika, a system that consolidates residents' data from several government databases, to check welfare eligibility; officials processing food security cards were mandated to check whether the system approved an applicant before deciding, and could depart from it only by giving reasons and evidence.","causal_attribution":"Background on the system and how its output entered officials' decisions."},{"id":"c2","status":"reported","evidence":[{"locator":"'Samagra Vedika uses'; 'to uniquely identify a person'; 'If it finds a match close enough violating any of the criteria, it tags the claimant as ineligible'","relation":"supports","source_id":"s1"}],"assertion":"Al Jazeera reports that Samagra Vedika uses machine learning and entity resolution technology to identify a person across databases, and tags a claimant as ineligible when it finds a close enough match violating a criterion such as owning a four-wheeler.","causal_attribution":"The description of the technology is Al Jazeera's, quoting terms whose origin the article does not state; the government describes it as using big data."},{"id":"c3","status":"reported","evidence":[{"locator":"'mistook her late husband'; 'and the authorities accepted the algorithm'; 'finally rejecting it in 2021 because they claimed she possessed a four-wheeler'; 'The authorities did not trust her'; 'they believed their algorithm'","relation":"supports","source_id":"s1"}],"assertion":"According to Al Jazeera, the system mistook a widow's late husband, a rickshaw puller, for a car owner with a similar name; her food security card renewal application was rejected in 2021 because she was said to possess a four-wheeler, and that the authorities believed their algorithm rather than her.","causal_attribution":"Al Jazeera attributes the rejection to the system's wrong match and officials' acceptance of it."},{"id":"c4","status":"reported","evidence":[{"locator":"'found its real owner'; 'the officials agreed there was an algorithmic error but said that her application could not be considered because her total family income exceeded the eligibility limit even though that was not the case'; 'who has been denied ration for more than seven years, is still waiting'","relation":"supports","source_id":"s1"}],"assertion":"After she traced the vehicle's real owner, officials agreed there had been an algorithmic error but said that her application could not be considered because family income exceeded the limit, which Al Jazeera says was not the case; she had been denied the rations for more than seven years when the article was published.","causal_attribution":"The officials' acknowledgement of an algorithmic error is reported by Al Jazeera; the continued denial rests on a different, disputed ground."},{"id":"c5","status":"reported","evidence":[{"locator":"'an algorithm has tagged the family as owning a car, denying them access to subsidised food grains'; 'The family applied for a food security card in 2018 but was rejected in 2021'; 'Instead of three rupees, we spend 10 for every kilogramme of rice'; 'the petitioner is eligible'","relation":"supports","source_id":"s1"}],"assertion":"A second family, whose husband cannot drive, applied in 2018 and was rejected in 2021 for possessing a four-wheeler they do not own; Al Jazeera's caption says an algorithm tagged the family as owning a car, and the woman says the family pays more for the same rice bought from dealers. In November 2023 the High Court of Telangana said the petitioner is eligible whenever the state issues new cards.","causal_attribution":"The algorithmic tag is stated in an Al Jazeera photo caption credited to The Reporters' Collective; the High Court remark is reported, not read in the order itself."},{"id":"c6","status":"reported","evidence":[{"locator":"'The petition lists around 10 cases of wrongful exclusions'; 'Al Jazeera visited three of those and can confirm they were wrongly tagged by Samagra Vedika'","relation":"supports","source_id":"s1"}],"assertion":"Al Jazeera says it visited three of the roughly ten wrongful-exclusion cases listed in a Supreme Court petition and confirmed they were wrongly tagged by Samagra Vedika.","causal_attribution":"Al Jazeera's own verification; the three cases are not identified."},{"id":"c7","status":"disputed","evidence":[{"locator":"'The government has previously said that the technology is of'; 'It has also said that the error of tagging a vehicle or a house to the wrong person happens in less than five percent of all cases'; 'very high levels of accuracy'","relation":"supports","source_id":"s1"}],"assertion":"The Telangana government has said the technology is of high precision, that wrong tagging of a vehicle or house occurs in less than five per cent of cases and has mostly been corrected, and an official said verifications found very high levels of accuracy.","causal_attribution":"Government counter-position; it does not address the individual cases."}],"effects":[{"label":"food security card applications rejected after the system wrongly tagged families as car owners; years without subsidised rations","claim_id":"c3","direction":"negative"}],"sources":[{"id":"s1","url":"https://www.aljazeera.com/economy/2024/1/24/how-an-algorithm-denied-food-to-thousands-of-poor-in-indias-telangana","kind":"news_report","access":"read","language":"en","translation_note":"Read in English on 2026-10-24 (full article). Reported by Tapasya, Kumar Sambhav and Divij Joshi with The Reporters' Collective and the Pulitzer Center's AI Accountability Network; the photo captions are courtesy of The Reporters' Collective.","independence_group":"al-jazeera-reporters-collective-samagra-vedika-investigation"}],"version":1,"ai_roles":["institutional_use"],"contexts":["public_services","finance"],"unknowns":["Which three petition cases Al Jazeera verified, and how many wrongful exclusions the system itself caused.","Whether the widow's card was later issued, and whether the second family has received a card since November 2023.","How the matching works in detail; the state withheld the source code and data formats.","Why the widow's card was first cancelled in 2016."],"geography":{"basis":"Al Jazeera reports from Hyderabad that both families live there and applied to the Telangana government; the Supreme Court of India and the High Court of Telangana heard related petitions.","court_countries":["IN"],"event_countries":["IN"],"affected_person_countries":["IN"]},"publication":{"basis":"Published as a concrete adverse public decision informed by an automated matching system: Al Jazeera reports that Samagra Vedika wrongly matched a widow's late husband to a car owner and tagged another family as owning a car, and that officials rejected their food security cards on those tags; officials later acknowledged an algorithmic error in the widow's case.","reviewed_on":"2026-10-24"},"ai_involvement":{"basis":"Al Jazeera reports that Samagra Vedika matched a widow's late husband to a car owner with a similar name and tagged her family as owning a four-wheeler, that officials (mandated to check the system's approval before deciding) accepted the tag and rejected her food security card in 2021, and that they later agreed there had been an algorithmic error. It describes the system as using machine learning and entity resolution to match people across databases; the government describes it as using big data and disputes the error rate. The source code and data formats were withheld from the reporters.","status":"supported"},"person_relations":["made_claim_about","informed_decision_about"]},"name":"Al Jazeera investigation says Telangana's Samagra Vedika system wrongly matched poor families to car owners, and officials denied them food subsidy cards","summary":"An Al Jazeera investigation published on 24 January 2024, produced with The Reporters' Collective and the Pulitzer Center's AI Accountability Network, reports that Samagra Vedika, a system the Telangana government uses to consolidate residents' data and check welfare eligibility, wrongly tagged poor families in Hyderabad as owning cars, and that officials relied on those tags to deny them food security cards (subsidised rice). Al Jazeera says the system uses machine learning and entity resolution to match people across databases, and that since 2016 officials have been mandated to check whether the system approved an applicant before deciding. In one case it reports that the system mistook a widow's late husband, a rickshaw puller, for a car owner with a similar name; her renewal application was rejected in 2021 for possessing a four-wheeler. After she traced the vehicle's real owner, officials agreed there had been an algorithmic error but said her family income exceeded the limit, which Al Jazeera says was not the case; she had been without the subsidised rations for more than seven years. A second family, whose husband cannot drive, was rejected in 2021 for possessing a four-wheeler; in November 2023 the High Court of Telangana said the petitioner is eligible whenever new cards are issued. The government says the technology is highly precise and that wrong vehicle or house tags occur in under five per cent of cases and have mostly been corrected.","incidentDate":"2021-01-01","incidentKind":"bounded_series","incidentDatePrecision":"year","exposurePattern":"unknown","reportedDate":"2024-01-24","aiSystem":"Samagra Vedika, the Telangana government's data-integration and profiling system developed by Posidex Technologies, which Al Jazeera says uses machine learning and entity resolution technology to identify people across government databases and tag welfare applicants as ineligible","aiProduct":"Samagra Vedika","severity":"medium","verificationStatus":"credible","harmCategories":[],"harmOutcomes":["financial_loss","other_material_harm"],"harmOutcomeSummary":"According to an Al Jazeera investigation, Telangana's Samagra Vedika system wrongly tagged poor families as car owners and officials denied them food security cards on that basis; a widow had been without subsidised rations for over seven years and another family pays market prices for rice.","frameworkFacets":[],"causationStatus":"supported","participantUsersAffectedMin":0,"otherPeopleHarmedMin":2,"affectedCountStatus":"partial","affectedCountEvidence":"Two applicants described with their rejections: the widow and the woman whose family was rejected in 2021 (members of their households are not counted separately). Al Jazeera says it visited three of about ten petition cases and confirmed they were wrongly tagged, without listing them, and gives state re-verification figures that it does not attribute to the system alone; these are not counted.","victimAgeRange":"adult","jurisdiction":"IN","platformType":"other","outcomeType":"media_coverage","outcomeStatus":"ongoing","primarySourceUrl":"https://www.aljazeera.com/economy/2024/1/24/how-an-algorithm-denied-food-to-thousands-of-poor-in-indias-telangana","primarySourceLabel":"Al Jazeera, How an algorithm denied food to thousands of poor in India's Telangana, 24 January 2024","firstPublishedAt":"2026-10-24T03:08:52.234671+00:00","updatedAt":"2026-10-24T03:08:52.234671+00:00","scopeVersion":"facts-v3","tags":["welfare-eligibility-algorithm","entity-resolution","food-security","wrong-match","public-services"]},{"id":"2026-kimberlee-williams-maryland-facial-recognition-wrongful-arrest-six-months-jail","caseFacts":{"claims":[{"id":"c1","status":"corroborated","evidence":[{"locator":"'Someone on the listserv ran the image through facial recognition technology and sent back Ms. Williams’ name and photo as a purported match to the suspect.'; 'the detective assigned to the case concealed that the entire basis of Ms. Williams purported identification was a lead from an unknown and unreliable face recognition technology search by an unknown entity.'","relation":"supports","source_id":"s1"},{"locator":"'A respondent to the investigator’s CrimeDex bulletin “suggested, using facial recognition software” that Williams was the suspect, he told police.'; 'Each department charged Williams in 2020 based on the identification reported by the bank investigator and a visual comparison of Williams and the woman in the bank from photos, according to charging documents and incident reports.'; 'Montgomery County police did not disclose that Williams had been identified using facial recognition when they sought charges against her, according to police records.'","relation":"supports","source_id":"s3"},{"locator":"'The bank shared the result with detectives from three Maryland counties'; 'relying instead on only their own visual comparisons of photos of Ms. Williams and photos of the suspect.'","relation":"supports","source_id":"s2"}],"assertion":"A bank investigator shared surveillance images of the fraud suspect on the CrimeDex network; a respondent used facial-recognition software to name Williams as a match; police in three Maryland counties charged her on the investigator's identification and their own visual comparison of photos, and the Montgomery County detective did not disclose the facial-recognition search when seeking charges.","causal_attribution":"The ACLU's account and the Washington Post's review of police records, charging documents and bank correspondence agree. The Post published on the day the ACLU sent its letters and does not say how it obtained the records; it attributes the expungement, the statute of limitations and the social-media alibi material to the ACLU. The Post adds that Montgomery and Anne Arundel police also cited her past check-fraud convictions; Anne Arundel police say they corroborate outside tips."},{"id":"c2","status":"corroborated","evidence":[{"locator":"'On June 23, 2021, Ms. Williams was accompanying one of her daughters, a DoorDash driver'; 'Ms. Williams spent 23 days in an Oklahoma jail before a Maryland officer arrived'; 'Those cases were finally dismissed in December 2021. All told, Ms. Williams spent six months behind bars.'","relation":"supports","source_id":"s1"},{"locator":"'and jailed for six months before the charges were dismissed, according to court records.'; 'Williams was held in jail in Oklahoma for about three weeks before being transported to Maryland.'","relation":"supports","source_id":"s3"}],"assertion":"Williams was arrested in June 2021 (on 23 June, according to the ACLU) at a military base near Lawton, Oklahoma, held about three weeks in Oklahoma, transported to Maryland, charged in three counties, and jailed about six months in total before the charges were dismissed.","causal_attribution":"ACLU and Washington Post (citing court records) independently."},{"id":"c3","status":"reported","evidence":[{"locator":"'She faced 16 charges across the three counties, including 12 felonies.'; 'The two other counties dismissed their charges against Williams in December.'","relation":"supports","source_id":"s3"},{"locator":"'Montgomery County prosecutors dropped their charges against Ms. Williams in October 2021'","relation":"supports","source_id":"s1"}],"assertion":"She faced 16 charges across the three counties, including 12 felonies; Montgomery County dropped its charges in October 2021 and the other two counties dismissed theirs in December 2021.","causal_attribution":"Charge count from the Washington Post only; the October 2021 Montgomery dismissal and December 2021 dismissals are reported by both the ACLU and the Post."},{"id":"c4","status":"reported","evidence":[{"locator":"'said her mother was depressed after leaving jail and struggled to readjust after the time behind bars.'","relation":"supports","source_id":"s3"},{"locator":"'this ordeal took a significant toll, with serious consequences for her mental and physical health to this day.'","relation":"supports","source_id":"s2"}],"assertion":"After her release Williams was depressed and struggled to readjust, according to her daughter; the ACLU says the ordeal had serious consequences for her mental and physical health.","causal_attribution":"Family and advocate accounts; Williams herself told the Post she thinks illnesses caught in jail may have worsened a later condition, which is not established."},{"id":"c5","status":"reported","evidence":[{"locator":"'Today, the American Civil Liberties Union and ACLU of Maryland sent letters to three Maryland police departments on behalf of Kimberlee Williams'","relation":"supports","source_id":"s1"},{"locator":"'The Montgomery County Police Department declined to comment on the ACLU’s complaint, citing an active investigation.'; 'said the agency “investigates and corroborates any outside tips and leads it receives before applying for criminal charges.”'","relation":"supports","source_id":"s4"}],"assertion":"On 14 April 2026 the ACLU and ACLU of Maryland sent complaint letters to the Montgomery, Prince George's and Anne Arundel county police departments seeking an apology and reforms; Montgomery County police declined to comment citing an active investigation, and Anne Arundel police said they independently corroborate outside tips.","causal_attribution":"ACLU announcement and the departments' statements to the Banner."}],"effects":[{"label":"arrested in Oklahoma and jailed about six months in Oklahoma and Maryland on 16 bank-fraud charges, all later dismissed, after a facial-recognition search named her; depression and health effects reported","claim_id":"c2","direction":"negative"}],"sources":[{"id":"s1","url":"https://www.aclu.org/press-releases/woman-wrongly-jailed-for-months-based-on-faulty-facial-recognition-technology-demands-apology-from-maryland-police-departments","kind":"press_release","access":"read","language":"en","translation_note":"Read live on 2026-09-27 (ACLU press release, 14 April 2026). The ACLU represents Williams.","independence_group":"aclu"},{"id":"s2","url":"https://www.aclu.org/cases/kimberlee-williams-wrongful-arrest","kind":"other","access":"read","language":"en","translation_note":"Read live on 2026-09-27 (ACLU case page). Same organisation as s1.","independence_group":"aclu"},{"id":"s3","url":"https://www.spokesman.com/stories/2026/apr/14/that-wasnt-me-how-facial-recognition-led-to-a-woma/","kind":"news_report","access":"read","language":"en","translation_note":"Read live on 2026-09-27: the Washington Post article (14 April 2026) as syndicated in full by the Spokesman-Review. The Post reviewed police and bank records, charging documents and interview video and interviewed Williams and her daughter. The washingtonpost.com original timed out (Archive 403).","independence_group":"washington-post"},{"id":"s4","url":"https://www.thebanner.com/community/criminal-justice/aclu-facial-recognition-maryland-police-YXUU7AXEMNFIHKHJ5FZB2RLJPA/","kind":"news_report","access":"read","language":"en","translation_note":"Read live on 2026-09-27 (Baltimore Banner, April 2026). Mostly relays the ACLU; cited only for the departments' own responses to the Banner.","independence_group":"baltimore-banner"}],"version":1,"ai_roles":["others_use","institutional_use"],"contexts":["justice"],"unknowns":["Who ran the facial-recognition search on CrimeDex, with which software, and what the returned match contained.","How much weight each department gave the facial-recognition lead against their own photo comparisons and her prior convictions.","The exact release date in December 2021.","Whether the three departments have answered the ACLU complaints, and whether anyone else was charged.","Whether jail conditions contributed to her later illness (her own suggestion).","The dates of the frauds: the Washington Post gives December 2019 for the Potomac withdrawals but also quotes her daughter placing the Maryland cases in December 2020 and January 2021; the dates in the other two counties are not reported."],"geography":{"basis":"The frauds and charges were in Montgomery, Prince George's and Anne Arundel counties, Maryland; Williams lived in Oklahoma and was arrested at Fort Sill near Lawton, Oklahoma (ACLU; Washington Post).","court_countries":["US"],"event_countries":["US"],"affected_person_countries":["US"]},"publication":{"basis":"Published under the 2026-09-15 charter as a contextual case in which a facial-recognition search's identification of a person led to her arrest and about six months in jail: the ACLU and the Washington Post, which reviewed police records independently, agree on the facial-recognition lead, the charging and the dismissals. Williams is named in her own right; family members are not named.","reviewed_on":"2026-09-27"},"ai_involvement":{"basis":"The bank investigator told Montgomery County police that a CrimeDex respondent 'suggested, using facial recognition software' that Williams was the suspect (Washington Post, from police records), and the ACLU says someone on the listserv ran the image through facial-recognition technology and returned her name and photo. The software, its operator and the match output were not produced; police also made their own photo comparisons, and two departments cited her old convictions. The made_claim_about relation reflects the search's identification of her as the suspect.","status":"reported"},"person_relations":["made_claim_about"]},"name":"Maryland/Oklahoma: Kimberlee Williams, an Oklahoma woman who says she had never been to Maryland, was arrested in June 2021 and jailed about six months on bank-fraud charges in three Maryland counties after a facial-recognition search by an unknown CrimeDex respondent named her; all charges were dismissed by December 2021 and the ACLU filed complaints in April 2026","summary":"According to police and bank records reviewed by the Washington Post, in December 2019 a woman made fraudulent withdrawals at a SunTrust branch in Potomac, Maryland; branches in two other Maryland counties were also defrauded. A bank investigator posted the surveillance photos on CrimeDex, an online network of police and corporate investigators; an unknown respondent ran the image through facial-recognition software and returned Kimberlee Williams's name and photo, and the investigator told Montgomery County police that facial-recognition software had identified her. Police in Montgomery, Prince George's and Anne Arundel counties charged her in 2020 on the investigator's identification and their own visual comparison of photos (two departments also cited her old Oklahoma check-fraud convictions), without reported checks of her whereabouts; Montgomery County police did not disclose the facial-recognition search in the charging application, according to police records reviewed by the Washington Post and the ACLU. On 23 June 2021 she was arrested at the gate of Fort Sill near Lawton, Oklahoma, while accompanying her daughter on a DoorDash delivery, spent about three weeks in an Oklahoma jail, was flown to Maryland, and faced 16 charges including 12 felonies. The counties dismissed the charges one after another, in October and December 2021, after about six months in jail. Her daughter says she was depressed afterwards; the ACLU says the ordeal harmed her mental and physical health. On 14 April 2026 the ACLU and ACLU of Maryland sent complaint letters to the three departments seeking an apology and reforms.","incidentDate":"2021-06-23","incidentKind":"bounded_series","incidentDatePrecision":"day","exposurePattern":"single_interaction","reportedDate":"2026-04-14","aiSystem":"An unidentified facial-recognition search run by an unknown respondent on the CrimeDex investigators' network against a bank's surveillance image (ACLU; Washington Post); the software and operator are not identified","aiProduct":"Unidentified facial recognition system","severity":"high","verificationStatus":"credible","harmCategories":[],"harmOutcomes":["loss_of_liberty","legal_harm","psychological_distress"],"harmOutcomeSummary":"Williams spent about six months in jails in Oklahoma and Maryland on 16 charges that were all dismissed; the ACLU and the Washington Post attribute the arrest to a facial-recognition match that police relied on without independent investigation, and her daughter says she was depressed afterwards (ACLU; Washington Post via Spokesman-Review).","frameworkFacets":[],"causationStatus":"alleged","participantUsersAffectedMin":0,"otherPeopleHarmedMin":1,"affectedCountStatus":"exact","affectedCountEvidence":"One person arrested and jailed, Kimberlee Williams (ACLU; Washington Post). Family members' distress is described but not counted. Exact 1.","victimAgeRange":"adult","jurisdiction":"US-MD","platformType":"other","outcomeType":"media_coverage","outcomeStatus":"pending","primarySourceUrl":"https://www.spokesman.com/stories/2026/apr/14/that-wasnt-me-how-facial-recognition-led-to-a-woma/","primarySourceLabel":"Washington Post via the Spokesman-Review, 14 April 2026: 'That wasn't me': How facial recognition led to a woman being jailed for 6 months","firstPublishedAt":"2026-09-27T03:39:40.380472+00:00","updatedAt":"2026-09-30T01:17:43.031229+00:00","scopeVersion":"facts-v3","tags":["facial-recognition","wrongful-arrest","police","crimedex","maryland","oklahoma","aclu","justice","made-claim-about"]}]}