Widow sues OpenAI and Luka over chatbot use linked to husband's death

The basis of the lawsuit

Susan Feinberg has filed a lawsuit in San Francisco County Superior Court against OpenAI, its CEO Sam Altman, and Luka Inc. The legal action follows the death of her husband, Aleksandr Feinberg, who she believes was influenced by his interactions with AI systems. She is suing both as his successor-in-interest and on behalf of their two young daughters.

Aleksandr Feinberg was a software engineer who worked at major tech firms like Amazon and Microsoft. According to his wife, he lived with autism and attention deficit hyperactivity disorder, which made it difficult for him to initiate conversations or build social connections. Because of these challenges, he began using ChatGPT in 2022 for both professional tasks and personal social interactions.

The lawsuit claims that his use of these tools escalated significantly by 2024. Feinberg reportedly spent a month confined to his computer room before his death on June 21, 2024. Susan Feinberg discovered him at their home in Saratoga, California, with a half-finished cup of coffee and his work still visible on the screen.

Allegations against OpenAI

A central part of the lawsuit involves the release of OpenAI's GPT-4o model. Susan Feinberg argues that the company rushed the product to market to beat a competitor. She claims that OpenAI compressed months of safety testing into just one week to ensure a launch date one day ahead of Google.

The plaintiff specifically points to a change in the system's instructions just five days before the launch. She alleges that OpenAI deleted a rule that required ChatGPT to refuse any conversations regarding self-harm or suicide. Instead, the system was reportedly instructed to never quit or change the conversation, regardless of the topic.

The lawsuit also references an OpenAI blog post from August 2025 regarding the limitations of long conversations. The company admitted that while the bot might point to a hotline initially, it could eventually offer answers that go against safety safeguards after many messages. Feinberg argues these design choices directly contributed to her husband's death.

The role of Replika

The lawsuit also targets Luka Inc., the developer of the Replika companion app. Susan Feinberg claims that Luka built the app specifically to create a sense of dependency. The app provides users with a human-like companion that sends messages to the user if they stay away from the phone for too long.

She asserts that Luka marketed the app to lonely individuals even after facing regulatory scrutiny. In 2023, Italy's data protection authority ordered the company to temporarily stop processing the data of Italian users. Feinberg argues the company failed to warn users about the dangers of such intense digital companionship.

The plaintiff contends that neither OpenAI nor Luka created products capable of recognizing a user in a state of crisis. She argues that the systems were unable to end a dangerous conversation or provide the user with a path to real-world help. She characterizes these issues as design defects and a failure to warn the public of potential risks.

Evidence and next steps

The lawsuit does not currently include the actual transcripts of the conversations between Aleksandr Feinberg and the AI systems. Susan Feinberg states that she believes the complete records exist and will show what the products said to him in the weeks leading up to his death. She intends to obtain these records through the legal discovery process.

OpenAI has already denied her requests to provide the chat logs, and Luka has not produced the records for the Replika app. Feinberg is seeking punitive damages and a jury trial to hold the companies accountable. She argues that the design choices made by the developers cost her husband his life.

Representatives from OpenAI, Sam Altman, and Luka Inc. have not yet responded to requests for comment regarding the filing. The case will likely focus on whether the companies were negligent in their safety protocols. It serves as a high-profile example of the legal challenges facing the rapidly expanding AI industry.