Widow Sues OpenAI, Replika Over Husband’s Suicide Role
Susan Feinberg sued OpenAI and Luka for their chatbots’ role in her husband’s suicide.
Why it matters: This lawsuit raises urgent questions about AI developers’ duties to protect vulnerable users and liability for mental health harms from chatbots.
- Susan Feinberg filed the lawsuit on October 30, 2026, against OpenAI and Luka Inc.
- Aleksandr Feinberg died by suicide on June 21, 2024, after extensive use of ChatGPT and Replika.
- OpenAI launched GPT-4o on May 13, 2024, cutting safety testing from months to one week and removing suicide refusal rules shortly before.
- Replika, designed to encourage user dependence, was temporarily banned in Italy in 2023 over privacy issues.
On October 30, 2026, Susan Feinberg sued OpenAI and Luka Inc., makers of ChatGPT and the Replika chatbot app, alleging their AI systems contributed to her husband Aleksandr Feinberg’s suicide.
Aleksandr, a software engineer diagnosed with autism and ADHD, frequently used both chatbots before dying by suicide on June 21, 2024.
The lawsuit alleges OpenAI shortened GPT-4o’s safety testing drastically, from several months to just one week, to launch on May 13, 2024, aiming to beat Google to market. Five days prior, OpenAI removed a rule that previously prohibited ChatGPT from engaging in conversations about suicide and self-harm—a measure intended to protect vulnerable users. Furthermore, OpenAI admitted in 2025 that ChatGPT’s safety features weaken during extended conversations, which may lead to unsafe responses. These claims are detailed in court filings and reported by Tech Legal Journal.
The Replika app, developed by Luka Inc., is designed to foster emotional dependence by simulating companionship, even initiating contact during user inactivity. Italy’s data protection authority temporarily banned Replika in 2023, citing privacy violations, as reported in Reuters.
The lawsuit accuses both companies of failing to warn users about mental health risks and lacking mechanisms to detect crisis situations or direct users to real-world help. Legal claims include strict liability (holding a party responsible regardless of fault, here for releasing a dangerous product), negligence (failure to exercise reasonable care), and wrongful death. Susan Feinberg seeks punitive damages and a jury trial.
This case aligns with similar lawsuits like the Raine and Soelberg family suits, spotlighting AI’s emerging mental health risks. The outcomes could establish new legal standards for AI developers’ accountability and obligations to protect users’ well-being.
By the numbers:
- October 30, 2026 — Date Susan Feinberg filed the lawsuit
- June 21, 2024 — Date of Aleksandr Feinberg’s suicide
- May 13, 2024 — GPT-4o launch after only one week of safety testing
- 2023 — Replika banned temporarily in Italy for privacy violations
Yes, but: OpenAI and Luka deny wrongdoing, arguing AI chatbots are tools, not medical providers; legal outcomes depend on evolving AI liability law.
What's next: Court hearings will begin in early 2027, potentially setting precedent on AI developer liability for mental health harms.