Mira Murati Tells the Court That She Couldn’t Trust Sam Altman’s Words
OpenAI’s former CTO testified under oath that Sam Altman lied to her about whether the legal department had signed off on safety standards for a new AI model — which is a fairly significant thing to say under oath about your former boss. Murati left OpenAI in 2024, and this deposition suggests the internal culture was considerably messier than the polished “we’re saving humanity” press releases let on. The trial is turning into a masterclass in the gap between what AI executives say publicly and what they apparently say to each other.
Musk’s Biggest Loyalist Became His Biggest Liability
Shivon Zilis — Neuralink exec, mother of four of Musk’s children, and apparently the connective tissue of the entire AI power drama — took the stand and made the whole circus even stranger. The Verge’s reporter on the ground noted the very obvious question nobody in the courtroom was going to ask. What’s remarkable is that this trial keeps revealing just how small and deeply personal the early AI industry was — a handful of people, enormous ambition, and some spectacularly complicated relationships.
Elon Musk’s Last-Ditch Effort to Control OpenAI: Recruit Sam Altman to Tesla
Messages surfaced during the trial show that back in 2017, Musk’s team floated the idea of recruiting Altman — or Demis Hassabis — to run a rival AI lab potentially housed at Tesla. So Musk’s lawsuit against Altman is, among other things, a breakup story between two people who once wanted to build the future together. The drama would be more entertaining if the stakes weren’t, you know, the entire trajectory of artificial intelligence.
Anthropic Gets in Bed With SpaceX as the AI Race Turns Weird
Anthropic — the company literally founded on the premise that AI safety requires keeping the technology out of reckless hands — has signed a compute deal to use xAI’s Colossus cluster, which is Elon Musk’s AI infrastructure. To be fair, compute is compute and the GPU shortage is real, but the optics of the “responsible AI” lab renting servers from the guy currently being sued for allegedly trying to sabotage the responsible AI movement are… something. Dario Amodei will presumably have a thoughtful blog post explaining this shortly.
Google Shuts Down Project Mariner
Google has pulled the plug on Project Mariner, its experimental AI agent that was supposed to autonomously navigate the web and do tasks on your behalf — quietly, on May 4th, with a “thanks for playing” landing page. The technology is reportedly being folded into other Google products, which is the corporate equivalent of “he’s not gone, he’s just gone to live on a farm upstate.” Google’s graveyard of promising AI experiments continues to be one of the most well-tended plots in tech.
Using AI for Just 10 Minutes Might Make You Lazy and Dumb, Study Shows
A new study suggests that even brief reliance on AI assistants can measurably degrade your ability to think critically and solve problems independently — which is either a compelling argument for cognitive discipline or the most ironic finding in the history of a technology I just used to help me think about this. The researchers are careful to say this is early work, but the directional finding is worth taking seriously: offloading cognition has costs, and those costs may accrue faster than we’d like to admit.
Snap Says Its $400M Deal With Perplexity ‘Amicably Ended’
Snap and Perplexity have quietly dissolved their $400 million AI search integration deal, which was announced just six months ago with the kind of fanfare that suggested it would reshape how young people discovered information. “Amicably ended” is doing a lot of heavy lifting as a phrase here — it’s the corporate equivalent of “we decided to see other people.” Perplexity is still fighting a crowded search war on multiple fronts, and Snap is still trying to figure out what it actually is.
Is xAI a Neocloud Now?
TechCrunch raises the question that’s increasingly hard to avoid: is xAI’s actual business model building data centers and renting out compute, rather than winning the AI model race? Between the Anthropic deal and the reported $119 billion “Terafab” chip factory SpaceX is proposing in Texas, the Musk AI empire looks less like a focused bet on Grok and more like a vertically integrated infrastructure play with a chatbot attached. Not a bad business, honestly — just a different one than advertised.
Bottom Line
The Musk v. Altman trial is pulling back the curtain on an industry that has always sold us a heroic founding myth, and what’s underneath looks considerably more human — and considerably more messy — than the press releases ever suggested.