Musk Trial Reveals OpenAI's Hassabis Obsession
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Greg Brockman testified that Elon Musk talked about Hassabis "many, many times" in OpenAI's early years, once asking at an early AI dinner, "Is Demis Hassabis evil?"
- In a 2016 email to Brockman and Ilya Sutskever, Musk wrote that DeepMind was "playing the Super Bowl and we are playing the Puppy Bowl" — language he used to argue for more urgency at OpenAI.
- Hassabis warned Musk in January 2016 that open-sourcing AI was "actually very dangerous," a message Musk forwarded to Altman and Sutskever shortly after OpenAI's founding.
- Shivon Zilis, then an OpenAI board member, wrote a personal plea asking Musk to "slow down" Hassabis, calling it "the only nonnegotiable net good action I can see."
- By November 2018, Musk had "lost confidence" OpenAI could counterweight Google/DeepMind, writing: "Unfortunately, humanity's future is in the hands of Demis."
- Zilis told Musk the OpenAI inner group met in a London coffee shop without cell phones to discuss strategy away from Hassabis, who they feared "would spy on their email and gchat."
- The last Hassabis mention in trial exhibits is from March 2019, when Sam Altman texted Musk that he had "some mild Demis updates to share."
Why it matters: The exhibits show that Musk's longstanding fear of Hassabis drove existential decisions inside OpenAI — including proposals to fold the lab into Tesla and Brockman/Sutskever's 2017 warning that Musk himself could become an "AGI dictator" if he seized more control. The competitive anxiety toward a single rival visibly shaped the trajectory of what is now the world's most-watched AI company.
