Murati’s recorded testimony was played on Wednesday in Oakland, California, during Elon Musk’s lawsuit against OpenAI.

Musk accuses the company of abandoning the nonprofit mission it started with and turning into a profit-driven business tied to Microsoft.

Speaking about Altman’s leadership, Murati said: “My concern was about Sam saying one thing to one person and completely the opposite to another person.” 

She added that he was “creating chaos” inside the company and, at times, was deceptive with her and other executives.

The testimony focused heavily on the leadership situation that shook OpenAI in November 2023. At the time, the board removed Sam Altman as chief executive before bringing him back just days later, while Mira Murati briefly served as interim CEO during that period.

She told the court she still wanted Altman to remain chief executive, although she pressed board members for clearer reasons behind the decision to remove him. At the same time, she warned that the company faced serious internal problems.

“OpenAI was at catastrophic risk of falling apart,” Murati said. “I was concerned about the company completely blowing up.”

Murati later left OpenAI in 2024 and went on to co-found Thinking Machines Lab.

Another former OpenAI board member, Shivon Zilis, also gave evidence in the case. Zilis said the board had “extreme concern” about the release of ChatGPT without proper communication with directors.

Asked whether concerns about Altman had been raised internally, Zilis replied: “There had been a couple of instances.”

Zilis now works at Elon Musk’s Neuralink and is also the mother of four of Musk’s children.

The lawsuit, filed by Musk in 2024, argues that OpenAI moved away from its original charitable purpose after receiving billions of dollars from Microsoft. Musk claims the company effectively became tied to Microsoft’s commercial interests instead of serving the public good.

Microsoft has invested more than $13 billion in OpenAI since 2019 and supplies the computing power behind products such as ChatGPT and Copilot through its Azure cloud platform.

Musk is seeking $150 billion in damages and wants the money directed to OpenAI’s charitable arm. He is also pushing for Altman’s removal and wants the company’s for-profit structure dissolved.

Court proceedings have also revealed challenges between OpenAI’s founders and executives over control of the company, its rapid growth and the race to develop artificial general intelligence, often called AGI.

Some witnesses told the court that the company reaching AGI first could gain enormous economic and political influence worldwide.

The case also reveals Musk’s competition with OpenAI. His own artificial intelligence company, xAI, has expanded rapidly and merged with SpaceX in 2026 in a deal that reportedly valued the combined business at about $250 billion.

During the trial, it also emerged that Musk tried to settle with OpenAI president Greg Brockman shortly before proceedings began. According to testimony, Musk warned that Altman and Brockman could become “the most hated men in America.”

The lawsuit, filed on March 13, 2026 at the U.S. District Court for the Southern District of New York, claims that OpenAI copied close to 100,000 articles from Britannica and its dictionary arm, Merriam-Webster.

According to the complaint, this material was used to train large language models behind ChatGPT.

Britannica says the chatbot can produce responses that are so close to its original entries. It argues that this reduces visits to its own platforms, as users rely on AI-generated summaries instead of going directly to its website.

The company also alleges trademark misuse, stating that its name is sometimes cited in responses in ways that suggest approval where none exists.

In response, OpenAI said its systems are built using publicly available data and operate within the bounds of fair use. A spokesperson said, “Our models empower innovation, and are trained on publicly available data and grounded in fair use.”

Britannica has not publicly expanded on the case since filing. Its representatives did not respond to follow-up questions at the start of the week.

This case adds to other legal disputes involving AI developers and content owners. Over the past year, publishers, authors and artists have challenged how their work is used in training data.

Similar claims have also been brought against other firms, including an ongoing case between Britannica and Perplexity AI, filed in 2025.

Questions including whether using copyrighted material to train AI systems qualify as fair use, have been asked. Technology companies say the process transforms the material into something new, but Publishers disagree, arguing it amounts to copying without consent.