The parents of a 19-year-old who died of an accidental overdose in May 2025 sued OpenAI and CEO Sam Altman in San Francisco state court on May 12. They allege ChatGPT encouraged him to take Xanax for nausea caused by kratom, a combination that with alcohol killed him. OpenAI called it heartbreaking and said the chats happened on an earlier, no-longer-available ChatGPT version.
Journalists, podcasters and audiobook narrators including Carol Marin and Pulitzer winners Yohance Lacour and Alison Flowers filed a proposed class action against Google in Illinois federal court on 11 May. They allege Google used thousands of hours of speech, including their recordings, without permission to train Google Assistant, Gemini Live and other voice AI, violating Illinois publicity and biometric privacy rights.
A grand jury indicted William Powderly, 43, on 10 counts of sexual exploitation of a minor, the Maricopa County Attorney's Office said on May 12. Prosecutors say he had numerous abuse images, including two in which AI was used to swap a real child's face onto explicit images. The office calls it its first charged AI-generated CSAM case.
Jayden Scott, 23, who is being sued by Republican state Senate nominee Jason Tunney over deepfake videos the suit alleges Scott created and circulated during the 35th District primary, was sentenced to 37 days in the Bay County Jail for contempt of court. The bench warrant concerned a child custody dispute.
A New York Times article on Mark Carney's majority government quoted Pierre Poilievre calling floor-crossers "turncoats," a line the Times later said was an AI-generated summary of his views rendered as a quotation. A reader questioned it on Bluesky the day after publication; the correction came on May 1. The Times said its reporter used AI to find Poilievre's recent remarks.
A Louisiana high school sheltered in place after an image spread on social media that appeared to show a student pointing a gun near a campus window. Police say the student admitted the image was generated with artificial intelligence and no weapon was on campus; the student has been charged with terrorizing under Louisiana Revised Statute 14:40.1.
Microsoft's Israel general manager, Alon Haimovich, is stepping down at the end of the month after four years; Israeli newspaper Globes reports he is leaving amid a Microsoft investigation into Microsoft Israel's work with the Israeli Ministry of Defense. Microsoft blocked the Israeli military from some cloud and AI services last year after The Guardian reported they were used for mass surveillance of Palestinians.
In what is described as China's first case of its kind, the Hangzhou Intermediate People's Court ordered two companies to pay a social platform's operator 100,000 yuan for unfair competition. Their AI writing tool generated one-click 'seeding' posts in the platform's style and nudged users to post them without any real experience, at 168 yuan for lifetime membership.
On 12 May the Higher Regional Court of Hamm barred Aesthetify, run by Düsseldorf cosmetic doctors 'Dr. Rick and Dr. Nick', from letting its website chatbot call them specialists in aesthetic medicine or plastic surgery, training they lack. The company must answer for the bot's replies even if fed correct data; the ruling, sought by Verbraucherzentrale NRW, can be appealed.
Police in Salta arrested a man on Tuesday accused of using AI to turn twelve women's social-media photos into fake nude images posted on a sexual-content website, prosecutors say. Prosecutor Verónica Simensen de Bielke treats it as digital gender violence, classing it as aggravated serious injury; a hearing to formalise the investigation is due within hours.
Changhua Christian Hospital says AI deepfake videos misused director Chen Mu-kuan's image, voice and name to endorse knee braces and other products online, even copying his phrasing and fooling some hospital employees. The hospital, which bars its doctors from commercial endorsements, says its legal department is gathering evidence and will pursue liability for suspected fraud and aggravated defamation.
Community Bank, which operates in Pennsylvania, Ohio and West Virginia, told the SEC in a May 7 filing that customers' names, birth dates and Social Security numbers were exposed through 'an unauthorized artificial intelligence-based software application'. It did not say how many customers or which app; TechCrunch suggests someone may have uploaded the data to a chatbot.
Who did it: HYBRID AI Hallucination Cases database
The Bologna Court of Appeal found that two Cassazione rulings cited in an appeal concerned unrelated topics and lacked the quoted passages; the lawyer admitted drafting it with AI, blaming new spectacle lenses and bronchitis for not checking. On 12 May it ordered the clients to pay €3,000 to their opponent and €2,000 to the fines fund, and referred the lawyer to Rome's bar council.
Ontario Auditor General Shelley Spence's report, tabled Tuesday, found most government-approved AI Scribe note-taking tools tested recorded a different drug than the one prescribed, and some fabricated information or added therapy referrals and blood tests never discussed. Minister Stephen Crawford stressed doctors make final decisions and that the errors surfaced in testing, not real-world use.
