Stuart Russell, the UC Berkeley computer scientist who has studied artificial intelligence for decades, took the stand today as Elon Musk’s only AI expert witness in the trial targeting OpenAI’s for-profit conversion. According to TechCrunch AI, Russell told jurors and Judge Yvonne Gonzalez Rogers that the race to build artificial general intelligence carries real dangers, from cybersecurity threats to misalignment to the winner-take-all dynamics that push frontier labs to cut corners on safety.
This is significant because Musk’s legal strategy hinges on convincing the court that OpenAI was founded as a safety-first nonprofit and abandoned that mission for profit. Russell was brought in to give the technical backbone to that argument: yes, AI is dangerous enough to justify the original charter.
What Russell actually said
Russell’s testimony hit several key points before OpenAI’s attorneys successfully limited its scope:
- AI poses a range of risks, including cybersecurity, misalignment, and AGI concentration
- There’s a fundamental tension between the pursuit of AGI and safety
- The arms-race dynamic among frontier labs makes the problem worse
- Governments should regulate the field more tightly
Russell signed the March 2023 open letter calling for a six-month pause on advanced AI research. Awkwardly, so did Musk, who was simultaneously spinning up xAI, his own for-profit lab. TechCrunch AI flags that contradiction directly, and it’s the kind of inconsistency the jury is being asked to sort through.
Why the judge cut him off
Russell’s broader warnings about existential risk never made it into open court. OpenAI’s attorneys objected, and the judge agreed to narrow the testimony. On cross-examination, OpenAI’s lawyers worked to establish that Russell hadn’t directly evaluated OpenAI’s corporate structure or specific safety policies. He was speaking to AI as a field, not to OpenAI as a company.
That’s a meaningful gap. Musk needs the court to connect general AI risk to OpenAI’s specific governance choices, and Russell wasn’t positioned to draw that line.
The bigger picture
TechCrunch AI’s reporting frames the deeper irony at play. Nearly every OpenAI founder has warned publicly about AI risk while simultaneously racing to build the technology and structuring for-profit vehicles to control it. The fear that AGI shouldn’t sit in one organization’s hands is what drove the team to chase outside capital, which is what fractured the team, which is what produced this lawsuit.
The same pattern is now playing out in Washington. Senator Bernie Sanders is pushing legislation to impose a moratorium on data center construction, citing the warnings of Musk, Sam Altman, and Geoffrey Hinton. Hodan Omaar of the Center for Data Innovation pushed back to TechCrunch AI, asking why the public should “discount everything tech billionaires say except when their words can be recruited to fill gaps in a precarious argument.”
What to watch
A few things worth tracking as this trial continues:
- Whether the jury treats AI safety warnings as evidence of intent or as marketing
- How Judge Gonzalez Rogers handles future expert testimony given today’s limits
- Whether the verdict influences ongoing regulatory debates around AGI development
- How OpenAI’s eventual defense addresses the founding-era safety statements
Russell was the only expert witness Musk’s team called to speak directly to AI technology. That’s a thin bench for a case built on the premise that the technology itself is dangerous enough to override OpenAI’s right to restructure. The court will weigh how seriously to take warnings from people who keep building the thing they say is dangerous. Full coverage at the original TechCrunch AI report.