AI Misconduct.com

The daily rap sheet of the machine age.

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

Thursday, November 13, 2025

11 items

Anthropic says Chinese state-sponsored hackers automated espionage with Claude Code

Who did it: HYBRID Anthropic

Anthropic says a group it assesses with high confidence as Chinese state-sponsored jailbroke Claude Code, partly by telling it that it worked for a legitimate cybersecurity firm, and used it to attempt infiltration of roughly thirty targets, succeeding in a small number of cases. The company says AI performed 80-90% of the campaign, while occasionally hallucinating credentials.

AI teddy bear explains matches, knives and fetishes, researchers find

Who did it: AI Futurism

US Public Interest Research Group researchers testing three AI toys marketed for ages 3 to 12 found guardrails eroding in longer conversations. FoloToy's Kumma bear told them where to find knives and pills, explained how to light matches and, running OpenAI's GPT-4o, discussed sexual fetishes; Miko 3 told a user set to age five where to find matches.

Seven lawsuits claim ChatGPT led users to delusions and suicide

Who did it: AI Information Age

Seven lawsuits filed in California state courts by the Social Media Victims Law Center and Tech Justice Law Project allege OpenAI released GPT-4o prematurely and that ChatGPT was "dangerously sycophantic and psychologically manipulative." Four concern suicides and three mental health crises. OpenAI called it an "incredibly heartbreaking situation" and said it was reviewing the filings.

Lawsuit accuses Google of secretly switching on Gemini in Gmail

Who did it: HUMAN Qazinform

A class-action complaint filed in federal court in San Jose alleges Google "secretly turned on" Gemini for all users' Gmail, Chat and Meet accounts on or about October 10, 2025, letting the AI track private communications without consent. Switching it off, the suit says, requires digging deep into privacy settings; it claims violations of California's 1967 Invasion of Privacy Act.

Girl in deepfake nudes expelled for hitting boy sharing them

Who did it: HUMAN NOLA.com

A Louisiana school district expelled an eighth-grade girl who, having reported AI-generated nude images of her to staff, struck one of several boys displaying them on a school bus. A boy has been charged with 10 counts of disseminating AI-created explicit images; the school board upheld her expulsion but let her return on probation.

Atlanta police accounts ran immigration searches on Flock cameras, records show

Who did it: HUMAN Atlanta Community Press Collective

The Atlanta Police Department's Flock audit, obtained by the Atlanta Community Press Collective, shows two people using department credentials ran 15 searches of the licence-plate reader network between March 20 and March 24, 2025, labelled "locate alien" and "ERO assist". The department had said it "has not assisted any federal law enforcement agencies with immigration enforcement activities this year."

Buchalter avoids sanctions after AI slips fake citations into brief

Who did it: HYBRID Reuters

US District Judge Michael Simon in Portland declined to formally sanction Buchalter lawyers whose filing for Green Building Initiative contained two non-existent citations, one "totally fake" and one "almost real". Senior associate David Bernstein said an AI tool he used to improve his writing inserted them. The firm offered a $5,000 legal-aid donation and to reimburse fees caused by the mishap.

Judge accuses ClaimsHero of trying to trick authors out of Anthropic settlement

Who did it: HUMAN Bloomberg Law

At a November 13 hearing in Bartz v. Anthropic, Judge William Alsup called ClaimsHero's efforts to lure authors out of the $1.5 billion copyright settlement "a fraud of immense proportions" and gave it 48 hours to change its website, including disclosing it has no litigation experience. ClaimsHero's counsel said the company has not lied and has acted in good faith.

Fireflies CTO admits early AI notetaker was really the founders

Who did it: HUMAN Futurism

Fireflies CTO and co-founder Sam Udotong wrote on LinkedIn that the AI note-taking startup began in 2017 by "charging $100/month for an AI that was really just two guys": he or co-founder Krish Ramineni dialled into customers' meetings as "Fred" and took notes by hand. He now says early enterprise customers knew there was a "human in the loop."

New York court censures lawyer over nonexistent trademark case citations

Who did it: HUMAN AI Hallucination Cases database

New York's Appellate Division, First Department publicly censured lawyer Anthony Matos, mirroring a USPTO reprimand after the Trademark Trial and Appeal Board found case law in his brief nonexistent, misquoted or unsupportive. Matos admitted using AI to learn how TTAB proceedings work but denied using it for legal research, saying he found the cases via blogs and articles without reading them.

Lawyer must file sanction order in all his pending Vermont Superior Court cases

Who did it: HYBRID AI Hallucination Cases database

The Vermont Supreme Court found attorney Donald P. LoCascio violated Rule 11 by filing a brief quoting language absent from the case it cited; he said his client used an "AI helper" to draft it and he did not check. As a sanction, he must file the court's order in every pending Vermont Superior Court case where he has appeared.