17 Apr 2026 to 31 May 2026Event location unknownMeta AI support assistant
Meta announced the rollout of its AI support assistant on Facebook and Instagram on 19 March 2026. Meta's filing with the Maine Attorney General (notice dated 5 June 2026) says unauthorized third parties exploited a vulnerability in its AI-assisted Instagram account recovery tool (High Touch Support) to receive password reset links for accounts they did not own, and the listing gives 17 April 2026 as the breach date and 31 May 2026 as the discovery date. Videos that attackers posted, as described by TechCrunch, 404 Media and Krebs on Security, show attackers asking the assistant in a chat to link a new email address to a target username. Reported victims include the security researcher Jane Manchun Wong, who posted that her password was changed without her knowledge, the Instagram account of the U.S. Space Force's Chief Master Sergeant, and the accounts of Sephora and a dormant Obama White House page (TechCrunch marks the last as disputed by Meta). Krebs and the BBC report pro-Iran defacement of some accounts, and TechCrunch reports that some victims were locked out and that short handles were offered for resale. The filing gives 20225 persons affected in total and 30 in Maine, and the notice calls the Maine figure an upper bound. Meta says the tool worked as intended, that a bug in a separate code path failed to check the email address, and (through a spokesperson to Gizmodo) that the failure was not due to the AI agent itself. Meta says it disabled the tool on 31 May 2026, the day it discovered the exploitation.
Contextual tracker case Low reported severity
AI involvement supported · Causal attribution disputed · 12 sources, 2 underlying accounts · Added 29/09/2026
20 May 2026United StatesMeta internal AI systems (reported)
On 20 May 2026 Meta began notifying about 8,000 employees (roughly 10 percent of its workforce) that they had been selected for layoff. On 13 July 2026 twenty-six anonymous employees who had taken or requested medical, pregnancy, parental or family leave, or a disability accommodation, sued Meta in the US District Court for the Northern District of California. The complaint alleges, on information and belief, that Meta used internal AI-assisted systems (including the Metamate assistant, keystroke and activity monitoring, AI-token-usage dashboards and algorithmic performance ranking) to score, rank and select employees, and that these inputs could not accumulate during protected leave, so plaintiffs on leave were disproportionately selected. Meta says workforce decisions were made by people, not AI, and that no selection decision was made by AI. On 17 July 2026 the court denied a temporary restraining order, recording that the parties dispute whether Meta used AI in the terminations and finding serious questions on the merits but no shown likelihood of success. One plaintiff was voluntarily dismissed on 3 August 2026. The preliminary injunction motion was argued on 24 August 2026 and taken under submission. The plaintiffs are pursuing their merits claims in arbitration and the allegations are unproven.
Contextual tracker case Medium reported severity
AI involvement disputed · Causal attribution disputed · 10 sources, 5 underlying accounts · Added 29/09/2026
Aug 2025Event location unknownUnidentified AI-text detector
In an August 2025 first-person post, a student said they discovered a zero for a summer-course discussion assignment after a checker labelled 85% of the text AI-generated. They denied using AI to write it, reported contacting the professor, and described anger and fear of future accusations. Neither the detector output nor the school’s account was available for verification.
Contextual tracker case Low reported severity
AI involvement reported · Causal attribution alleged · 1 source · Added 29/09/2026
21 Sept 2026 to 25 Sept 2026FrancePangram AI-text detector
On 21 September 2026 an anonymous X account, Balance ton Claude, said it had run excerpts of Thélyson Orélien's debut novel C'était ça ou mourir through the American AI-detection software Pangram, with some passages scoring up to 100 per cent probability of AI generation. On 25 September the Académie Goncourt removed the novel from the first selection of sixteen books for its 2026 prize, stating that the work is 'in all likelihood very largely the product of an artificial intelligence', based on 'the convergent result of several analyses by researchers and journalists', and also citing plagiarism in older texts by the author published in Canada (at least one short story and articles), which do not concern the novel. Orélien, a 38-year-old Canadian-Haitian writer whose novel had won the Fnac novel prize, denies using AI and says its repetitions and certain rhythms reflect his linguistic and cultural world and may be misread by automated tools. His Montreal publisher Boréal suspended promotional activities, saying it could neither confirm nor refute the allegations. Essayist Samuel Fitoussi said he was one of the people behind the account. Whether AI was used to write the novel is disputed; AI-detector reliability is itself contested in the coverage.
Contextual tracker case Medium reported severity Internal Action
AI involvement reported · Causal attribution disputed · 4 sources, 1 underlying account · Added 28/09/2026
19 Jun 2026 to 13 Aug 2026BrazilChatGPT and OpenAI moderation
Over about two months in 2026 a 36-year-old farmer in rural São Gabriel da Palha, Espírito Santo, exchanged messages with ChatGPT that the Espírito Santo civil police describe as plans to kill his eight-year-old son, attack other people and kill himself; one message said he had offered 50,000 to someone to kill him and his son. OpenAI reported the messages to the FBI, which passed them through Brazil's Ministry of Justice cyber-operations laboratory to the state police on 16 June; the man was arrested on 19 June as he left home, a day before the date on which police believed the plan would be carried out. Police say what the platform provided was corroborated at the scene (four bottles of unidentified substances and a knotted rope were found) but that he denied intending to kill the child; his lawyer says he was talking nonsense to a chatbot and took no concrete step. After 54 days in custody a court granted habeas corpus on 13 August 2026 because the inquiry had not been concluded and no charges had been brought, imposing electronic monitoring and a ban on approaching or contacting his son and the child's mother. Police were still examining his phone and awaiting forensic results. OpenAI provided user details and his messages but not ChatGPT's replies and did not answer the BBC's question why; it told the BBC it may notify law enforcement when it detects a credible and imminent risk of harm to others. The case is recorded as an institutional response to AI use: the reported adverse consequence is the detention without charge, and the police account that the report prevented a planned killing is preserved as context.
Core + contextual relations High reported severity Investigation Opened
AI involvement supported · Causal attribution supported · 4 sources, 2 underlying accounts · Added 23/09/2026
2 Sept 2026United StatesMeta AI
On 1 September 2026 Kalie Robins, a Utah travel creator, posted a short Instagram video of herself singing in the car with one of her daughters; it was also shown on Facebook. The next day she noticed Meta AI had placed suggested questions under the post, starting with 'Who's the child passenger?'. Clicking the prompts, she said, produced her two daughters' names, birth information, photos and videos drawn from her own and relatives' past posts, including a newborn photo from her mother's account and a picture she believed she had deleted years earlier, and a further prompt, 'Where does Kalie Robins live?', assembled older and newer posts into her likely location. Her 2 September reaction video drew more than 310,000 likes. Meta told reporters the prompts 'missed the mark', 'should never have been generated' and had been fixed, while saying the feature only surfaces information the user can already access; it disputed that a long-deleted photo was used, saying the photo had been deleted shortly before the video and remained briefly visible through a bug. Robins said Meta never contacted her and that the episode left her feeling she had 'failed' her children.
Contextual tracker case Low reported severity Involving minors Internal Action
AI involvement supported · Causal attribution supported · 4 sources · Added 18/09/2026