I’ve been keeping an eye on the various documents released as part of the ongoing Musk vs Altman legal fight, which centres around OpenAI changing to a for-profit structure and Elon Musk’s belief he was deceived by Sam Altman. The documents, which consist mainly of emails and text exchanges between the various figures involved, turn out to have several themes running through them: and an absolutely huge one is the co-founder and CEO of Google DeepMind, Demis Hassabis.
Hassabis is widely regarded as the outstanding talent in the AI field, a fellow of the Royal Society, and in 2024 shared the Nobel Prize in Chemistry with John M. Jumper for their work on AI protein structure prediction. He also began his career in videogames at Bullfrog, before working as lead AI programmer on Lionhead’s Black & White, and founding his own developer, Elixir Studios (which made Republic: The Revolution and Evil Genius).
One intriguing message from Zilis to Musk [PDF] has, unfortunately, been redacted, though again shows how he’s always uppermost in their thoughts: “Will leave you be on the Demis stuff. I’m sure it’s hard to think about [redacted]. I just needed to say it once since it’s been plaguing me. [redacted].”
A few days later Musk is back on his Hassabis hobbyhorse.
“OpenAI is not a serious counterweight to DeepMind/Google and will only get further behind. It is surprising that this isn’t obvious to you,” writes Musk. “In general, always overestimate competitors. You are doing the opposite.”
But Hassabis also weighed heavily on the minds of others. Satya Nadella gave pre-trial testimony in September 24, 2025 [PDF], and one of the lawyer’s questions is “Google was—I believe you said by far the dominant player in machine learning around 2015?”
“I would say so, yeah,” replies Nadella, before the lawyer mentions Google’s acquisition of DeepMind and asks whether Microsoft believed the company was “making a lot of progress in that field of machine learning?”
“I now can’t recall specifically what state DeepMind’s breakthroughs were, but yeah DeepMind was well known, even in that time frame, and Google had DeepMind, had Google Brain. They had many, many different efforts they were publishing.”
Nadella’s testimony can be condensed into his acknowledgement that Google led the AI race from the 2010s onwards, and that Microsoft was tracking DeepMind’s progress. Nadella says “I probably knew [Hassabis] a little” from around 2015 “just after I became CEO.”
The lawyer drills down and asks why Nadella and Microsoft were tracking DeepMind specifically.
“Just because of the breakthroughs that this particular regime of AI around deep neural networks were showing real promise of making breakthroughs in fields like language translation that had not been seen before. And so that’s why we were waiting to see how we could also participate and make sure that we have those breakthroughs.”
Amusingly enough, the lawyer moves on to OpenAI and Dota 2. One of OpenAI’s first goals was to create a bot that could beat humans at Dota 2, a task in which it eventually succeeded: And as PCG reported last week, that was because Elon Musk had personally called Microsoft CEO Satya Nadella to secure a massive discount on access to Azure, the company’s cloud computing platform.
Here, the lawyer begins by asking what Nadella understands about the game.
“I’m not a gamer,” says Nadella. “And I think it’s a Steam game, if I’m not mistaken.” He then goes on to give a brief overview of why he thought the game angle with AI was interesting.
“That’s why gaming environments being closed worlds were a great sort of, you know, environment to do reinforcement learning,” says Nadella. “Right? The objective function is clear. The reward function is clear.
“I forget now when and what time frame some of the breakthroughs on AlphaGo and so on happened but, you know, Demis was a game developer. There’s a long history of AI developers who came out of using games in environments, building AI bots in games, so it’s sort of a given.”
Elon Musk also gave pre-trial testimony a few days later, on September 26, 2025 [PDF], in which he reveals that poaching people from DeepMind caused him to have a fallout with Google co-founder Larry Page.
“I talked to dozens and people over the years,” says Musk. “But I think the most crucial recruit was Ilya Sutskever. In fact, the recruitment of Ilya was what actually caused Larry Page to stop being friends with me.
“So Ilya went back and forth multiple times saying he would join OpenAI or stay at Google, and ultimately agreed to join; and Larry Page and Sergey [Brin] and Demis Hassabis did everything they could to keep Ilya. When Ilya finally decided to join OpenAI, that’s what ended the friendship with Larry Page. He didn’t talk to me after that.”
The lawyer asks if Musk is still getting the silent treatment a decade later, and it’s a simple “yes” before adding “they were very upset about [Sutskever].”
Let’s end on the funny note of Sam Altman’s delusions of grandeur. This document [PDF] was filed on 6 January 2026, but unfortunately is not otherwise dated. Based on what he’s saying though I would date it in the region of 2016-2018.
“Progress fundamentally has to be made by non-profit, interesting direction you could go,” writes Altman. “Everything I perceive with OpenAI, race dynamics vs Demis + brain + whatever, gotta get there first.”
I’m not quite sure but “brain” may be a reference to neural chips, though of course what’s amusing here is that once again Hassabis and the “race dynamics” of competing with him is living rent-free in Altman’s head. Then, the ego really takes over.
“Another angle: [DeepMind will] never do that much that’s interesting,” writes Altman. “It is better for us to become increasingly kings of this industry. The choice defines us.
“You say we should have been more ok with giving it to Elon. While it’s true, it’s also the case he’s now given it to us. The grand upside is I want it. Need to stop letting distractors get to me/us. Being the Kings of AI is not so bad.”
The Kings of AI! Good band name. Also: Jesus wept. Do we really want this dweeb potentially making big decisions about humanity’s future?
In early 2019 Altman is still trying to tempt Musk into phone calls by promising “some mild Demis updates to share” [PDF] while in 2023 Mira Murati is emailing Nadella [PDF] saying “it is very important that we don’t lose researchers to Demis or Elon.”
Let’s end on one of the least-crazy things anyone says about Hassabis, and it’s from Ilya Sutskever [PDF] who actually worked with the guy.
“The goal of OpenAI is to make the future good and to avoid an AGI dictatorship,” Sutskever writes to Musk in September 2017. “You are concerned that Demis could create an AGI dictatorship. So do we.
“So it is a bad idea to create a structure [for OpenAI] where you could become a dictator if you chose to, especially given that we can create some other structure that avoids this possibility.”
It’s a great point: but answer came there none.
Best gaming rigs 2026