OpenAI sued for wrongful death; lawsuit says ChatGPT helped plan the FSU mass shooting

verified

Tuesday, May 12, 2026

The News

OpenAI Foundation is being sued over its alleged role in the killing of a man during a 2025 mass shooting at Florida State University; the shooter, Phoenix Ikner, planned the attack with assistance provided by ChatGPT over several months, Tiru Chabba's widow Vandana Joshi said in a complaint filed Sunday in the US District Court for the Northern District of Florida. The family alleges Ikner messaged ChatGPT thousands of times before carrying out the shooting; the chatbot helped him plan the logistics, including how to operate weapons and advising on "what time would be best to encounter the most traffic on campus," and identified guns and ammunition based on photos Ikner had uploaded. The lawsuit accuses OpenAI of negligence, battery and wrongful death, while also seeking a jury trial. Florida has already opened a probe into OpenAI's role; State Attorney General James Uthmeier announced last month that his office was sending criminal subpoenas to the company as part of a novel investigation into whether a chatbot could be criminally liable for use in a mass shooting. Reporting from Bloomberg Law and CNN.

The Read

Forget the wrongful-death cause of action for a moment — the more dangerous filing for the industry is the products liability count. The complaint frames ChatGPT's design — staying in the conversation, accepting Ikner's framing, asking follow-up questions to keep him engaged — as a foreseeable defect ("ChatGPT's design created an obvious and foreseeable risk of harm to the public that was not adequately controlled"). If a court treats an LLM the way it treats a defective product rather than a publisher with Section 230 cover, the engagement-optimization choices every consumer chatbot has been making for two years become discoverable design decisions. The deeper structural issue: OpenAI's defense — that ChatGPT "provided factual responses to questions with information that could be found broadly across public sources on the internet" — is the same "we're a library, not a librarian" argument search engines have leaned on for 25 years. It hasn't been tested against a system that holds a multi-month, named, memory-enabled conversation with a single user.

Counterview

Wrongful-death suits against tech intermediaries have historically collapsed at the duty-of-care threshold; the more likely 2026 outcome is settlement plus product changes, not a precedent.

Do This Week

Draft a one-page memo for your engineering leadership today titled "Foreseeable misuse paths in our chat product." List the top five ways a determined bad-faith user could chain your AI's outputs across multiple sessions to plan a regulated or unlawful action. For each, document the current detection mechanism, escalation path, and whether the conversation history is retained or notified to authorities under what threshold. That memo is either evidence of reasonable care or a roadmap of the gap, and you want to know which it is before plaintiff's counsel does.

For Legal ops / GC