AI Misconduct.com

The daily rap sheet of the machine age.

Updated Nov 2, 2026
4,173 items in the archive
RSS feed

Friday, August 1, 2025

9 items

Windows Recall filter blocks 'My SS#' but not 'Soc. #'

Who did it: AI Tom's Guide

Testing by The Register found Microsoft's Windows Recall, which screenshots everything on a Copilot+ PC, still captures sensitive data in many cases despite its default "Filter sensitive information" setting. It snapped bank balance and deposit screens and a document with passwords, and while it filtered a Social Security number labelled "My SS#", it captured one labelled "Soc. #".

Jury finds Tesla partly liable for fatal Autopilot crash

Who did it: HYBRID Ars Technica

A federal jury in Miami found Tesla one-third responsible for a 2019 crash in which a Model S on Autopilot ran a stop sign at 62 mph and struck two people stargazing, killing one and severely injuring the other. It awarded $129 million in compensatory and $200 million in punitive damages; Tesla calls the verdict wrong and plans to appeal.

Lawyer ordered to pay $85,567.75 over AI-hallucinated filings

Who did it: HYBRID AI Hallucination Cases database

A federal judge in the Southern District of Florida ordered attorney James Martin Paul, sanctioned on July 15, 2025 over filings with hallucinated cases and fabricated quotations, to pay defendants $36,663.00 and $48,904.75 in fees and costs within 90 days. The court rejected his offer of $20,000 in installments, saying capping liability would only benefit 'serial hallucinators'.

Deepfake Guernsey chief minister pitches fake investment scheme

Who did it: HUMAN BBC

Fraudsters used AI-generated video and images of Guernsey's chief minister, Deputy Lindsay de Sausmarez, appearing to recommend an investment scheme, alongside fake local newspaper articles, police say. Guernsey Police said any posts featuring deputies requesting payment or investment are a scam; de Sausmarez said she hoped most people would quickly realise it is fake.

TikTok finds deepfake Mercedes Milá diabetes ad breaks no rules

Who did it: HUMAN elEconomista.es

Spanish presenter Mercedes Milá says an AI-manipulated video using her face and voice is promoting a product that promises miracles against diabetes. She says she reported it to TikTok, which concluded within 30 minutes that it did not breach community rules; the video has passed half a million views. Manipulating images to sell a health product is 'very serious', she said.

Bahamas lawyer referred to ethics committee over ChatGPT fake cases

Who did it: HYBRID AI Hallucination Cases database

The Supreme Court of The Bahamas found that three cases defence counsel Darell Taylor cited in a strike-out application did not exist; she told the court she had used ChatGPT. Justice J. Denise Lewis-Johnson rejected her distinction between 'speaking points' and submissions, declined to strike out the claim, ordered costs against the defendant and referred the matter to the Bahamas Bar Association's Ethics Committee.

Ontario judge adjourns litigant's motion over AI-invented case law

Who did it: HYBRID AI Hallucination Cases database

Ontario Regional Senior Justice E. Ria Tzimas adjourned a motion by self-represented defendant Nicolas Rewa, with substantial indemnity costs to Halton, after three of the four cases in his factum proved nonexistent as cited and the fourth beside the point. Rewa admitted using AI; the judge suggested one suspiciously on-point invented test may reflect deliberate manipulation, and gave a final warning.