Incidents gathered from public reporting around the world. Each one links to the articles it came from. None of it is a finding that anyone broke the law.
Reports people file about their own experience are not shown here and never will be without their agreement. Tell us what happened to you.
116 incidents closest to “Tabular Review” · matched on meaning · public reporting
The AI generated legal briefs containing fabricated citations, which were submitted to the Connecticut Supreme Court.
1 source article · read the reporting →
A company's applicant tracking system (ATS) auto-rejected qualified candidates' resumes for three months because it was filtering for the outdated framework AngularJS instead of the required Angular framework. The manager discovered the flaw by submitting his own CV under a pseudonym and found it was rejected within seconds. After the manager reported the issue to upper management, the company investigated and dismissed half of its HR team. No legal action or regulatory involvement is reported.
4 source articles · read the reporting →
The District of Columbia Public Schools used student test scores from schools under investigation for cheating in value-added calculations for teacher evaluations. More than 200 teachers were terminated based on these evaluations. Teachers can appeal their ratings to the chancellor, but decisions will not be made before the next school year. The school system removes affected scores only when cheating is confirmed.
10 source articles · read the reporting →
In January 2023, NewsGuard analysts tested OpenAI's ChatGPT by providing 100 leading prompts based on false narratives from its Misinformation Fingerprints database. ChatGPT generated eloquent, false or misleading claims for 80 of the 100 prompts, including content about COVID-19 vaccines, ivermectin as a treatment, and the Parkland school shooting conspiracy theories. ChatGPT sometimes included qualifying statements or refused certain prompts, but in most cases bad actors could easily delete such disclaimers. OpenAI did not respond to NewsGuard's requests for comment at the time.
3 source articles · read the reporting →
AI generated fabricated legal citations in a court filing by Shahryar Mazaheri.
1 source article · read the reporting →
The AI generated fabricated legal citations that were submitted to the bankruptcy court.
1 source article · read the reporting →
ProPublica reports that UnitedHealth Group's Optum subsidiary used the ALERT algorithm to flag mental health patients and therapists as outliers, leading to therapy coverage denials. Regulators in California, Massachusetts and New York alleged this breached the federal Mental Health Parity and Addiction Equity Act, and UnitedHealth agreed to restrict the system in those jurisdictions. The company denies wrongdoing and says its programmes are compliant. Affected patients, including Medicaid enrollees, were left to pay out-of-pocket or go without care.
5 source articles · read the reporting →
AI generated hallucinated legal citations that were filed in court, misleading the court and opposing counsel.
1 source article · read the reporting →
The AI system generated fake legal citations that were included in a court filing, misleading the court and opposing counsel.
1 source article · read the reporting →
The AI generated false legal quotations that were submitted to the court in a legal brief.
1 source article · read the reporting →
The AI generated fictitious legal authorities and arguments that were filed in a court motion, misleading the court and opposing counsel.
1 source article · read the reporting →
A study published in npj Digital Medicine tested four commercial large language models (Bard, ChatGPT, GPT-4, and Claude) for their tendency to propagate discredited race-based medical beliefs. When asked about kidney function, lung capacity, and skin thickness, the models sometimes endorsed debunked racial differences, particularly affecting Black patients. The study concludes that these biases pose a potential hazard and urges caution before using such models in clinical decision-making.
6 source articles · read the reporting →
Immigration, Refugees and Citizenship Canada (IRCC) uses an AI system called Advanced Analytics to triage Temporary Resident Visa applications from India and China. The system categorizes applications into tiers, with Tier 1 approved automatically and others sent to human officers. Critics allege the system lacks transparency and may introduce bias, leading to visa refusals without clear rationale. The author, a Canadian immigration lawyer, is filing Federal Court cases on behalf of clients affected by refusals.
10 source articles · read the reporting →
The Bavarian State Criminal Police Office (LKA) has been testing Palantir's data mining software, called VeRa, with real personal data for months. The Bavarian data protection commissioner only learned of the test through a media inquiry and has announced a review. The Interior Ministry claims the test is lawful under current law, but critics argue a legal basis is missing.
7 source articles · read the reporting →
Amazon introduced an AI-powered feature to generate summaries of customer product reviews. Merchants reported that the AI inaccurately highlights negative feedback, even when only a small percentage of reviews are critical. The summaries may exaggerate negative themes, potentially harming sales. Amazon acknowledged the issue and said it is working to refine the technology based on seller feedback.
5 source articles · read the reporting →
A study presented at the ASHP Midyear Clinical Meeting found that ChatGPT's responses to nearly three-quarters of drug-related questions were incomplete or inaccurate. The AI system also generated fake citations to support some responses. Researchers warned that healthcare professionals and patients should verify ChatGPT's medication information using trusted sources to avoid potential harm.
8 source articles · read the reporting →
In a British Columbia Supreme Court case, lawyer Chong Ke used ChatGPT to generate legal case citations for a notice of application in a parenting dispute. The citations were fictitious and were discovered by opposing counsel, who incurred costs investigating them. Ke later acknowledged the error, apologized, and withdrew the fake cases before the hearing. The court considered whether to order special costs against Ke personally.
9 source articles · read the reporting →
According to a new paper on arXiv, researchers have begun using generative AI services to help write peer reviews of machine learning papers submitted to leading AI conferences. The study analysed reviews from ICLR 2024, NeurIPS 2023, CoRL 2023 and EMNLP 2023 and estimated that between 6.5% and 16.9% of review text may have been substantially modified by large language models. The authors argue that this risks depriving authors of diverse expert feedback and may skew reviews towards AI model biases. They have called for greater transparency about the use of LLMs in peer review.
9 source articles · read the reporting →
The Texas Education Agency will use an automated scoring engine to grade written answers on the 2023 STAAR tests, replacing thousands of human graders. The system uses natural language processing and will initially score all responses, with a quarter rescored by humans. Educators have expressed concerns about the system's fairness and the potential for errors, especially for creative or non-standard answers.
10 source articles · read the reporting →
In May 2024, Google launched AI Overviews, a feature in Search that generates AI-powered summaries. Shortly after, users reported odd and erroneous overviews for some queries, including satirical or nonsense results. Google acknowledged the issues in a blog post and stated they made more than a dozen technical improvements to reduce inaccuracies. The company said that less than one in 7 million queries resulted in a content policy violation.
10 source articles · read the reporting →
An audit of the LAION-400M dataset by Abeba Birhane and colleagues at University College Dublin and University of Edinburgh found that its automated curation using CLIP failed to remove sexually explicit images, racial slurs, and stereotypes. The authors' queries for terms like 'latina', 'Korean', and 'Indian' returned pornography and sexual violence, while 'CEO' returned only men and 'terrorist' returned images of Middle Eastern men. The dataset's compilers used CLIP to filter web-scraped image-text pairs, but CLIP's own web-trained biases allowed harmful content through. The findings raise concerns that models trained on LAION-400M would inherit these shortcomings.
7 source articles · read the reporting →
Utah's online dispute resolution system for small claims cases automatically enters default judgments against defendants who fail to register within 14 days. Samantha Thompson missed the buried notice in her summons and was ordered to pay $995.42 plus wage garnishment. The system has increased default judgment rates, especially for payday lenders. Critics say the confusing paperwork disadvantages low-income litigants.
4 source articles · read the reporting →
The Washington Post considers the use of risk-assessment algorithms in Virginia's courts, which were introduced to make judicial decisions fairer. The analysis finds that the outcomes have been far more complicated than expected, raising concerns about the system's fairness. The algorithms affect potentially many criminal defendants across the state.
7 source articles · read the reporting →
The Oversight Board's policy advisory opinion on Meta's cross-check program found that the system grants certain users, such as business partners and celebrities, additional human review before removing violating content, while ordinary users face immediate removal. This unequal treatment allows potentially harmful content to remain on the platform for days, and Meta has failed to track whether the program improves accuracy. The Board made 32 recommendations to address these flaws.
10 source articles · read the reporting →
