Elon Musk described OpenAI’s Sam Altman as ‘so deceptive’ and suggested a 50 percent chance that humanity lives in a simulation during a conversation with Walter Isaacson on The Builders podcast.

Musk had been outlining a proposal for rival AI labs to examine one another’s models before launch when Isaacson asked whether he could discuss the arrangement with Altman. Meanwhile, his simulation estimate came from an anecdote about two dinner guests with opposing views.

Musk’s response to the prospect of speaking with Altman was blunt. ‘He’s so deceptive, it’s crazy.’ He did not identify a specific instance of deception in the quoted exchange.

Why Elon Musk Put Simulation Odds at 50 Percent
The Tesla and SpaceX chief described questioning his two dinner companions about the probability that they were living in a simulation. He recounted that their replies arrived at the same time.

He recalled that ‘the physicist said 0 percent and the computer scientist said 100 percent simultaneously. So I guess maybe I should go [with] 50 percent then.’

That figure was his suggested midpoint between their positions. The anecdote did not establish a scientific calculation of the odds, and Musk framed his reasoning about simulated existence conditionally.

He told Isaacson that ‘if simulation theory is correct, then any simulations that are boring will be truncated. Therefore, the fact that we exist means that we are an interesting simulation. And really the most important thing is just don’t be boring.’

Elon Musk Calls for Rival AI Safety Checks
On AI development, Musk proposed regular contact between major US labs and giving competitors a chance to examine new models before release. He envisioned weekly calls, with competitors granted access to an upcoming model a week before launch.

Musk suggested weekly calls among major AI labs and giving competitors access to new models a week before release, so they could spot potential problems before those models reached a wider audience.

‘I think competitors would keep each other honest in that situation. You know, if they say, oh, you know, I was able to make a nuclear bomb with your AI, it’s like, oh, we better fix that before we release it, you know, type of thing.’

The nuclear example illustrated the kind of problem he wanted those checks to detect. He did not describe an actual incident in that passage.

Read More: Elon Musk brands Sam Altman ‘so deceptive’ and claims we have a 50 percent chance of a simulation