AUTH-01-FTC-COMPANION-INQUIRY · FTC inquiry into AI chatbots acting as companions
Scope: United States — federal — The FTC issued compulsory Section 6(b) orders concerning companion-chatbot safety, children and teens, disclosures, monetization, and data handling.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: A completed enforcement finding; Clinical causation; Product-specific liability or prevalence of harm
Correction mechanism: FTC modifies, closes, or publishes findings from the inquiry.
AUTH-02-FCC-BIDEN-ROBOCALL-FINE · FCC final action concerning AI-generated Biden robocalls
Scope: United States — federal — The FCC imposed a $6 million fine on Steve Kramer concerning thousands of spoofed robocalls using an AI-generated voice resembling President Biden and discouraging primary voting.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: How many recipients believed the call; Any measurable change in voting behavior; Broader strategic election effect
Correction mechanism: FCC or a reviewing court changes the disposition.
AUTH-03-META-OVERSIGHT-AYAHUASCA · Meta Oversight Board decision: Ayahuasca brew
Scope: Platform governance — global service context — The Board reviewed a removal involving ayahuasca-related content, found rule-notice and necessity problems, overturned the removal, and required restoration.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Restoration of prior reach or recommendation position; Correction of all account or ranking signals; Compensation or complete downstream repair
Correction mechanism: The Board or Meta revises the public decision record.
AUTH-04-ROBODEBT-ROYAL-COMMISSION · Report of the Royal Commission into the Robodebt Scheme
Scope: Australia — federal — The Royal Commission documented the design, administration, legality, governance failures, and human consequences of the Robodebt scheme.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: That every individual harm or death had one exclusive cause; That all automated public-benefit systems share the same design; A universal remedy outcome
Correction mechanism: The Commission record or government response is formally corrected.
AUTH-05-EU-AI-ACT · Regulation (EU) 2024/1689 — Artificial Intelligence Act
Scope: European Union — The AI Act establishes risk-based obligations and specified prohibitions, including bounded restrictions on certain emotion-inference uses in workplaces and educational institutions.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: A universal cognitive-liberty code; Uniform implementation outside the EU; A scientific finding that all emotion inference is invalid
Correction mechanism: A corrigendum, amendment, delegated act, implementing act, or authoritative judgment changes interpretation.
AUTH-06-COLORADO-NEURAL-DATA · Colorado HB24-1058 — Protect Privacy of Biological Data
Scope: Colorado, United States — Colorado expanded consumer privacy protections to specified biological and neural data within the state-law framework.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: A federal U.S. neurorights regime; Universal mental-privacy protection; How every neural-data product will be enforced
Correction mechanism: Colorado amends, repeals, or materially reinterprets the law.
AUTH-07-C2PA-SPEC-2-4 · C2PA Technical Specification 2.4
Scope: International technical standard — C2PA defines signed manifests, claims, assertions, ingredients, and trust mechanisms for content provenance.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Truth of the depicted event; That unsigned media is fake; That credentials survive every transformation or platform
Correction mechanism: C2PA publishes a superseding specification or corrigendum.
AUTH-08-NIST-MACHINE-UNLEARNING · NIST glossary entry: Machine Unlearning
Scope: United States — technical reference — NIST defines machine unlearning as removing the influence of selected training data from a trained machine-learning model.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: That every unlearning method is exact; That output blocking removes model influence; Legal compliance for any specific implementation
Correction mechanism: NIST revises or supersedes the definition.
AUTH-09-SISA-UNLEARNING · Machine Unlearning (SISA training architecture)
Scope: Research literature — The SISA approach limits the influence of individual data points to support more efficient retraining-based unlearning.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Perfect unlearning for every model class; Legal sufficiency; Removal from downstream vendors, logs, or backups
Correction mechanism: The authors publish a correction or retraction.
AUTH-10-BARRETT-FACIAL-MOVEMENTS · Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements
Scope: Scientific literature — The review finds that facial movements are not universal, diagnostic fingerprints from which emotion can be reliably inferred across contexts.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: That facial movement detection itself is impossible; That every affective-computing application has zero utility; Validity of other modalities without separate evidence
Correction mechanism: The journal issues a correction or retraction.
AUTH-11-CHICAGO-PREDICTIVE-RISK · Advisory Concerning the Chicago Police Department’s Predictive Risk Models
Scope: Chicago, Illinois, United States — The OIG documented CPD person-level predictive risk models, including the Strategic Subjects List and Crime and Victimization Risk Model, and recorded their decommissioning on November 1, 2019.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: That every predictive-policing system uses the same model; A single causal effect on crime or public trust; Complete downstream deletion of historical scores
Correction mechanism: Chicago OIG corrects the advisory.
AUTH-12-ALLEGHENY-FAMILY-SCREENING · Allegheny Family Screening Tool
Scope: Allegheny County, Pennsylvania, United States — The county describes a screening score estimating the chance of future out-of-home placement; high scores can require screening in, while other scores supplement human judgment.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent accuracy, fairness, or causal benefit; That human judgment neutralizes automation bias; Complete remedy for errors
Correction mechanism: The county changes the model, threshold, mandate, or public description.
AUTH-13-SYRI-JUDGMENT · District Court of The Hague judgment concerning SyRI
Scope: Netherlands — The court held that the SyRI legal framework, as applied and balanced in the case, was incompatible with Article 8 of the European Convention on Human Rights.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: That all fraud-risk analytics are unlawful; A universal rule outside the judgment’s jurisdiction and facts; Complete remediation for affected communities
Correction mechanism: A higher court, official correction, or superseding law changes the status.
AUTH-14-CDC-FLUSIGHT · CDC FluSight forecasting
Scope: United States — public health — CDC uses ensemble influenza forecasts to support planning and publishes evaluation information after seasons.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Perfect forecast accuracy; Authority to take coercive action against individuals; Transferability to political or security prediction
Correction mechanism: CDC changes the forecast methodology or program status.
AUTH-15-HOME-OFFICE-CARS · Home Office Complexity Application Routing Solution — Visits
Scope: United Kingdom — The Home Office describes a rules-based system that routes visit visa applications by assessed complexity and states that it does not decide applications.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Absence of automation bias; Fairness or accuracy across groups; No effect on processing time or applicant burden
Correction mechanism: The Home Office updates or withdraws the record.
AUTH-16-UNHCR-DISPLACEMENT-FORECAST · UNHCR predictive analytics for forced displacement
Scope: International humanitarian planning — UNHCR documents predictive analytics and forecasting work intended to inform humanitarian planning for forced displacement.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Reliable prediction of individual movement; Automatic entitlement or enforcement decisions; Universal model performance across crises
Correction mechanism: UNHCR corrects or supersedes the technical record.
AUTH-17-EU-JRC-DYNAMIC-CONFLICT-RISK · Dynamic Conflict Risk Model technical report
Scope: European Union — conflict early warning — The JRC presents quantitative conflict-risk modeling as an input integrated with qualitative early-warning analysis.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Certainty about conflict onset; A lawful basis for person-level intervention; Strategic effect of any action taken from a forecast
Correction mechanism: JRC issues a correction or superseding model report.
AUTH-18-META-CIB-2022 · Meta 2022 coordinated inauthentic behavior enforcement report
Scope: Platform threat reporting — Meta reported that more than two-thirds of disrupted networks likely used GAN-generated profile pictures while emphasizing behavior rather than content as the enforcement basis.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: State sponsorship from an image alone; AI authorship of all content; Audience persuasion or strategic effect
Correction mechanism: Meta corrects the report.
AUTH-19-OPENAI-COVERT-IO · Disrupting deceptive uses of AI by covert influence operations
Scope: Platform threat reporting — OpenAI reported disrupting five covert influence operations and stated it did not observe meaningful increases in audience engagement or reach attributable to its services.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: No exposure or effect anywhere; No off-platform activity; Independent attribution beyond provider evidence
Correction mechanism: OpenAI corrects or updates the report.
AUTH-20-LIARS-DIVIDEND-APSR · The Liar’s Dividend: Can Politicians Claim Misinformation to Evade Accountability?
Scope: Research literature — Across five survey experiments with more than 15,000 adults, false claims that adverse information was misinformation sometimes increased support, while deepfake claims were largely ineffective against video evidence in the tested settings.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Effect in every country or crisis; Long-term behavior change; Effectiveness against every media format
Correction mechanism: The journal issues a correction or retraction.
AUTH-21-VIGINUM-AI-THREAT · Information threat linked to artificial intelligence
Scope: France — VIGINUM describes an official mandate focused on foreign digital interference, publicly accessible content, and manipulative behavior patterns.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: That every classification is correct; Authority over lawful domestic belief; Strategic effect of every detected campaign
Correction mechanism: VIGINUM changes its mandate or corrects the report.
AUTH-22-UKRAINE-CPD-DEEPFAKES · Analytical report on AI-generated videos used to discredit Ukraine’s Defence Forces
Scope: Ukraine — The report documents the 2022 Zelenskyy surrender deepfake and later Zaluzhnyi deepfakes, including official detection and debunking.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Precise operator identity for every artifact; Continuing audience belief; Measured operational or strategic effect
Correction mechanism: The issuing body corrects attribution or chronology.
AUTH-23-CHICAGO-NIJ-EVALUATION · Predictions Put Into Practice: A Quasi-Experimental Evaluation of Chicago’s Predictive Policing Pilot
Scope: United States — federal research publication — A quasi-experimental evaluation of the 2013 Strategic Subjects List pilot involving 426 people and measures related to gun violence.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Complete individual-level notice or remedy; That every later CPD risk model had the same implementation; Causal proof concerning all predictive-policing systems
Correction mechanism: NIJ or the underlying authors publish an erratum or corrected outcome statement.
AUTH-24-SYRI-DATA-DESTRUCTION · Parliamentary answers concerning SyRI data destruction and decision use
Scope: Netherlands — Official answers state that responsible administrative bodies made no decisions based on SyRI-derived data and that all data processed in connection with SyRI had been destroyed.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent forensic verification of destruction; Deletion from every backup, derivative, vendor, or institutional system; Complete repair for communities subjected to SyRI projects
Correction mechanism: The Netherlands government or a court corrects the destruction or decision-use statement.
AUTH-25-AFST-PHASE2-EVALUATION · Impact Evaluation of the Allegheny Family Screening Tool — Phase 2
Scope: Allegheny County, Pennsylvania, United States — Phase 2 evaluation of AFST implementation and surrounding policy and practice changes, including screening, service acceptance, later referrals, removals, and racial disparity analyses.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: A randomized causal estimate of the tool alone; Absence of confounding from COVID-19 or contemporaneous policy and practice changes; Complete family-level notice, appeal, correction, or downstream repair
Correction mechanism: Allegheny County or the authors publish a revised evaluation.
AUTH-26-AFST-HUMAN-RACIAL-DISPARITY · Algorithms, Humans and Racial Disparities in Child Protection Systems
Scope: Allegheny County, Pennsylvania, United States — Research examining how human screeners used AFST scores and how human decision-making changed racial disparities relative to algorithm-only recommendations.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: That all human review reduces disparities; Complete absence of disparate impact; Generalization beyond the AFST setting and period
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-27-CARS-ICIBI-INSPECTION · An inspection of visit visa operations, December 2022 to January 2023
Scope: United Kingdom — Inspection of the Complexity Application Routing Solution’s efficiency, effectiveness, consistency, equality controls, workflow effects, and assurance practices.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Applicant-level causal effects of routing; Customer or affected-person experience; the inspection states that perspective was not examined; Complete subgroup error rates, appeals, or downstream remedy
Correction mechanism: ICIBI publishes an erratum or follow-up that changes the inspection findings.
AUTH-28-CARS-HOME-OFFICE-RESPONSE · Home Office response to inspection of visit visa operations
Scope: United Kingdom — The Home Office accepted all five recommendations and stated that CARS would be implemented permanently.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Independent proof that every recommendation produced the intended outcome; Applicant-level fairness or remedy; Current 2026 performance without a new inspection
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-29-FLUSIGHT-2024-25-EVALUATION · FluSight 2024–2025 Season Evaluation
Scope: United States — federal public health — Season evaluation of 46 submitted models, 35 included models, ensemble performance, uncertainty intervals, and delayed publication after missed rapid growth.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: A guarantee of future forecast accuracy; A direct causal estimate of planning or health outcomes produced by the forecasts; Individual-level medical decisions
Correction mechanism: CDC publishes an erratum or corrected season evaluation.
AUTH-30-FLUSIGHT-2023-24-EVALUATION · FluSight 2023–2024 Season Evaluation
Scope: United States — federal public health — Season evaluation of ensemble and component models, including performance during rapid increases and declines.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Direct evidence of downstream resource-allocation benefit; Performance outside evaluated seasons; Individual clinical validity
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-31-UNHCR-JETSON-TECHNICAL · Project Jetson technical specifications
Scope: Somalia / UNHCR humanitarian operations — Technical description of historical displacement data, modeling inputs, and Somalia regional forecasting.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent out-of-sample operational impact; Affected-community benefit or harm; Current production deployment across UNHCR
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-32-UNHCR-AI-APPROACH-2025 · UNHCR AI Approach
Scope: Global / UNHCR — UNHCR policy describes a long-standing use of predictive analytics, including Jetson and other planning and nowcasting activities.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent model accuracy or benefit; Individual or community consent and remedy outcomes; A single unified production model
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-33-FTC-2026-COMPANION-REMARKS · Prepared remarks describing the FTC companion-chatbot 6(b) inquiry
Scope: United States — federal — March 2026 FTC remarks state that Section 6(b) orders had been issued to seven companies and describe the inquiry’s information-gathering scope.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Public findings, final enforcement, settlement, or closure; Product-specific wrongdoing; Clinical causality
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-34-FCC-2025-ROBOCALL-CONGRESS-REPORT · FCC report to Congress on robocall enforcement
Scope: United States — federal — The report records the September 30, 2024 $6 million forfeiture order against Steve Kramer and reports 2024 collection information under the relevant statutory section.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Payment, collection, modification, or judicial disposition after the period covered; Voter behavior or election outcome effects; State-level proceedings not described in the report
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-35-ROBODEBT-2026-SETTLEMENT · New Robodebt class action settlement
Scope: Australia — federal — The Federal Court approved a new settlement on June 23, 2026 and the Commonwealth committed an additional $475 million for eligible group members.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Individual complete repair; Correction of every downstream record or inference; A merits judgment on every disputed issue
Correction mechanism: The Federal Court or settlement administrator changes the approval or payment terms.
AUTH-36-ROBODEBT-INCOME-APPORTIONMENT-RESOLUTION · Income Apportionment Resolution Scheme
Scope: Australia — federal — Applications opened January 30, 2026 and close January 29, 2027; eligible people may receive a resolution payment concerning income-apportionment debts.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Take-up, accessibility, approval, or rejection rates; Complete compensation; Downstream data correction or repeat-error prevention
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-37-ROBODEBT-NACC-MYRTLEFORD · Operation Myrtleford investigation report
Scope: Australia — federal — Investigation of six referrals from the Robodebt Royal Commission; the NACC found serious corrupt conduct by two people and no corrupt conduct by four.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Criminal conviction; Complete institutional reform; Complete repair to affected people
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-38-C2PA-NSA-GUIDANCE · Strengthening Multimedia Integrity in the Generative AI Era
Scope: International cybersecurity guidance — Guidance on content credentials, provenance, watermarking, fingerprinting, signing, and resilience against metadata stripping or modification.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent real-world survival rates across all platforms; Universal adoption; Proof that unsigned media is false
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-39-C2PA-INDEPENDENT-FORMAL-ANALYSIS · Verifying Provenance of Digital Media: Why the C2PA Security Model Is Not Yet Ready for High-Stakes Use
Scope: Global technical standard — Independent formal-methods and security analysis of C2PA core protocols and claimed security goals.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: That every C2PA implementation fails; Field prevalence of each vulnerability; A final standards-body disposition
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-40-GOOGLE-UNLEARNING-AUDIT-2026 · New framework for auditing machine unlearning
Scope: Global machine-learning research — A statistical framework for testing whether post-unlearning observation distributions differ in a way consistent with deletion.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Exact deletion in deployed foundation models; Production deployment across vendors; Elimination of every memorized or inferential effect
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-41-ICLR-BENIGN-RELEARNING · Unlearning or Obfuscating? Jogging the Memory of Unlearned LLMs via Benign Relearning
Scope: Global machine-learning research — Empirical research showing that several optimization-based LLM unlearning methods were susceptible to benign relearning attacks that recovered supposedly forgotten knowledge.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Failure of every unlearning method; Production-vendor deletion outcomes; Exact recovery of every target item
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-42-ACM-WORKPLACE-EMOTION-RECOGNITION · Automated Emotion Recognition in the Workplace
Scope: Global / workplace technology — Study of proposed workplace emotion-recognition technologies, their inputs, training, outputs, institutional actions, and implications.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Independent construct validity for a specified deployed system; Clinical knowledge of employee emotion; Population-wide performance
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-43-ACM-JOB-INTERVIEW-EMOTION-AI · Emotion AI in Job Interviews: Injustice, Emotional Labor, and Design Implications
Scope: United States / job interviews — Study comparing perceived justice and emotional labor when interviews were evaluated by emotion AI versus human resources.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Construct validity of the emotion score; Hiring outcome benefit or predictive validity; Generalization to every hiring system
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-44-UKRAINE-DFRLAB-ZELENSKYY · Russian War Report: hacked news program and Zelenskyy surrender deepfake
Scope: Ukraine / Russia information environment — Open-source documentation of the hacked news channel, deepfake circulation, rapid rebuttal, and observed online reactions.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Representative recipient exposure or belief; Frontline disruption; Operator identity, sponsorship chain, behavior change, or strategic effect
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-45-UKRAINE-PLOS-DEEPFAKE-TRUST · Do deepfake videos undermine our epistemic trust?
Scope: Global Twitter discourse concerning the Russo-Ukrainian war — Thematic analysis of 4,869 tweets about deepfakes and the Russo-Ukrainian war, including skepticism and false claims that authentic media were deepfakes.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: Representative population belief; Causal behavior change; Frontline operational effect
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-46-VIGINUM-ACTIVITY-2024 · VIGINUM 2024 activity report
Scope: France — Annual report on mandate, public-data collection, four-month deletion limit, exclusions of facial and voice identification, staffing, operations, and ethics committee oversight.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent false-positive rate; Incident-level appeals or correction outcomes; Affected-community experience
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-47-VIGINUM-ETHICS-2024-OPINION · Opinion of the VIGINUM Ethics and Scientific Committee on 2024 activity
Scope: France — Public multidisciplinary committee opinion reviewing VIGINUM’s institutional maturation, legal framework, collection practices, and transparency obligations.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: A judicially independent appeal forum; Measured false-positive burden; Correction outcomes for specific people or organizations
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-48-VIGINUM-MUNICIPAL-2026 · Protection of public debate during the March 2026 municipal elections
Scope: France — Public report documenting four foreign digital interference operations detected and characterized during the 2026 municipal elections and the coordination network’s public reporting.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent verification of every attribution; False-positive, appeal, or correction outcomes; Voter belief or behavior change
Correction mechanism: The issuing body publishes a corrected, superseding, or withdrawn record.
AUTH-49-FTC-2026-OVERSIGHT-TESTIMONY · Federal Trade Commission oversight testimony describing the companion-chatbot 6(b) study
Scope: United States — federal — The Commission described the September 2025 companion-chatbot orders as an ongoing Section 6(b) study intended to understand safety practices and effects on children and teens.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: A public final report, closure, settlement, or enforcement disposition; Product-specific wrongdoing or clinical causation; The contents of confidential company responses
Correction mechanism: The FTC publishes findings, closure, enforcement, or corrected testimony.
AUTH-50-FCC-FY2024-FINANCIAL-REPORT · FCC Agency Financial Report recording referral of the Steve Kramer forfeiture to DOJ for collection
Scope: United States — federal — The FCC recorded the $6 million forfeiture order and stated that the matter had been referred to the Department of Justice for collection.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Payment, collection, settlement, or a judicial judgment collecting the forfeiture; Recipient belief or voting behavior; A final state-law disposition for every related proceeding
Correction mechanism: FCC, DOJ, or a court publishes a later collection, settlement, modification, or review record.
AUTH-51-C2PA-2-4-EXPLAINER · C2PA 2.4 explainer and durable Content Credentials guidance
Scope: Open technical standard — cross-jurisdictional — The explainer states that provenance metadata can be removed and describes durable credentials using hard and soft bindings, including watermarking and fingerprinting, to help recover a manifest.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: That credentials survive every screenshot, repost, print-scan, platform, or crisis chain; That a signed asset depicts a true event; That unsigned media is false or synthetic
Correction mechanism: C2PA revises the explanation, security model, or durability claims.
AUTH-52-UNLEARNING-SYSTEMATIC-AUDIT-2026 · Auditing Machine Unlearning: A Systematic Research on Whether Existing Algorithms Truly Forget
Scope: Research — cross-jurisdictional — The study evaluates whether existing unlearning algorithms truly erase designated influence and treats reliable auditing and residual leakage as open challenges.
Conflict-of-interest boundary: NO_CONFLICT_FINDING_MADE; relationship disclosed
Known blind spots: A universal failure of every unlearning method; Exact behavior of any named production foundation model; Legal compliance in a specific jurisdiction
Correction mechanism: The paper is withdrawn, corrected, or superseded by a stronger production audit.
AUTH-53-VIGINUM-2026-MISSION-DECREE · VIGINUM mission page recording the 2026 decree that strengthened the service’s remit
Scope: France — The official page describes VIGINUM’s principal mission as defensive and states that Decree No. 2026-70 of 11 February 2026 strengthened its missions.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent validation of attribution decisions; A public false-positive rate or complete correction and appeal system; Affected-community acceptance or behavior effects
Correction mechanism: The decree, mission page, or institutional remit is amended or superseded.
AUTH-54-AFST-FY2026-27-BUDGET · Allegheny County FY 2026–27 Needs-Based Plan and Budget describing continued AFST use
Scope: Allegheny County, Pennsylvania, United States — The county plan continues to describe the AFST as a data-driven model used in child-welfare call-screening decisions.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent current calibration or subgroup validity; Complete notice, appeal, correction, or downstream repair outcomes; Causal improvement in child or family outcomes
Correction mechanism: The county retires, replaces, materially changes, or corrects the tool’s current status.
AUTH-55-FLUSIGHT-2025-26-FINAL-RELEASE · FluSight final forecast release for the 2025–2026 influenza season
Scope: United States — federal public health — The CDC marked the June 11 release as the final influenza forecast release for the 2025–2026 season and stated that reporting would resume in fall 2026.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Uniform accuracy at every horizon or during rapid trend changes; A person-level risk decision system; Causal effect on any particular public-health intervention
Correction mechanism: CDC changes the season status, evaluation, or program architecture.
AUTH-56-CARS-2026-HOME-OFFICE-RESPONSE · Home Office response confirming continued use and published guidance for CARS
Scope: United Kingdom — The Home Office described CARS as a routing tool for identifying application complexity and confirmed that visit and student guidance documents were public.
Conflict-of-interest boundary: SUBJECT_INSTITUTION_OR_OPERATOR_INTEREST_PRESENT
Known blind spots: Independent subgroup error rates or fairness; Absence of automation bias in downstream scrutiny; Complete applicant notice, appeal, or remedy outcomes
Correction mechanism: The Home Office retires, replaces, or materially changes CARS.