Telsa CEO Elon Musk spent part of his final day testifying in an Oakland courtroom Thursday hearing a federal judge warn him not to talk about “extinction” and his fears of artificial intelligence killing humanity.

While taking the stand as part of a billion-dollar civil case against OpenAI and its CEO, Sam Altman, Musk often clung to an apparent commitment to AI safety, at one point testifying that he met with then-President Barack Obama about the booming technology in 2015. Though Musk said Wednesday that he believes “some futures are good and some are not good” in reference to AI’s increasing prevalence in society, he said Thursday that he’s usually optimistic regarding the future of humanity.

But repeating a similar sentiment from Wednesday, he said the “worst-case situation” would be that “AI kills us all.” His remark prompted Judge Yvonne Gonzalez Rogers to intervene.

“You made your little statement, and that’s OK, but you are instructed not to talk about extinction again,” she said.

During his three days in court, Musk accused the company’s founders, including Altman, of “stealing a charity.” Musk, who’s also a founder, alleged that the OpenAI founders betrayed him by shifting it from nonprofit to for-profit, abandoning the “core mission” to which he donated $38 million. He testified that part of the motivation behind founding OpenAI was to create an AI lab focused on safety and to balance out Google’s early AI efforts.

Before Musk and the jury had even entered the courtroom Thursday, Gonzalez Rogers told attorneys that “catastrophe and extinction” shouldn’t be discussed, according to reporters in the courtroom. (SFGATE reported from the overflow room of the courthouse.) The judge said although she believes there are many people who do not want to “put the future of humanity in Mr. Musk’s hands,” the trial is not going to get into those topics, CNN’s Hadas Gold reported.

“This is not a trial on whether or not AI has risks,” Gonzalez Rogers said. “This is not a trial about whether or not AI has damaged humanity. That may be a trial in the future.”

Later in the day, the judge directly told Musk not to talk about “extinction” after his attorney, Steve Molo, questioned him about “the future.” Molo asked Musk if Tesla is making an AI “military army of robots,” a topic that was also referenced Wednesday.

Musk replied that Tesla does not create any weapons.

“If we make a lot of robots, we’ve got to make sure that they are safe and do not turn into a ‘Terminator’ situation,” he said.

One of the biggest arguments from OpenAI’s attorneys is that safety isn’t Musk’s biggest priority and that he only sued OpenAI in 2024 after creating his own AI company in 2023, xAI. During the fiery cross-examination on Tuesday, William Savitt, OpenAI’s lead attorney, even asked if xAI creates a safety risk, which Musk answered affirmatively and said, “All AI companies suffer from that.”

Musk’s testimony ended shortly after he accused Savitt of “conflating timelines” and asking him “leading” questions.

“That’s not how it works, Mr. Musk. Let’s remind everyone in the courtroom that you’re not a lawyer,” Gonzalez Rogers said.

“Well, technically, I did take Law 101,” Musk said.

“You’re not a lawyer,” Gonzalez Rogers reaffirmed, with the exchange prompting laughs from the courtroom and overflow room.

In a post on X shortly before the trial, OpenAI called Musk’s lawsuit “a baseless and jealous bid to derail a competitor.”

Altman and OpenAI co-founder and president Greg Brockman are set to testify later in the trial. After his testimony, Musk was allowed to leave for the day, but the judge said he is still “not excused.”

The trial comes just six months after OpenAI floated the idea of allowing it to go to public markets. SpaceX, another one of Musk’s companies, also announced plans to go public.