All the evidence revealed so far in Musk v. Altman
The trial is a gift that keeps giving — early emails, corporate documents, and internal exchanges from OpenAI’s pre-naming days are now part of the public record. The picture emerging is of a messy, ambitious founding that doesn’t quite match Musk’s tidy “I built this, they stole it” narrative. Grab the popcorn; there’s more to come.
Elon Musk Had a Bad Week in Court
Musk demanded this fight, spent months positioning himself as the wronged visionary behind OpenAI, and then apparently had a rough few days on the witness stand where his own tweets and emails were used against him. There’s a lesson here about the risks of suing a company when your entire paper trail is public — but I suspect it won’t be learned. When your strongest argument is “I wanted it to stay nonprofit” and you spent the last decade building a competing for-profit AI lab, the jury tends to notice.
How Shivon Zilis Operated as Elon Musk’s OpenAI Insider
Trial exhibits show Zilis — Neuralink exec, mother of four of Musk’s children, and OpenAI board observer — served as a direct communication channel between Musk and OpenAI leadership during the critical early years. This is the kind of detail that reframes the entire “Musk was just a donor” vs. “Musk was the driving force” debate, and it’s the sort of thing that makes this trial genuinely interesting beyond the celebrity beef.
Disneyland Now Uses Face Recognition on Visitors
The Happiest Place on Earth is now also the Most Surveilled Place on Earth — Disney has deployed facial recognition on park visitors, which is either a sensible crowd management tool or a deeply unsettling development depending on your tolerance for a corporation biometrically cataloging your family vacation. Buried in the same news cycle: the NSA is testing Anthropic’s Mythos Preview for vulnerability research, which is a sentence that would have sounded like science fiction three years ago.
Brace for the Patch Tsunami: AI Is Unearthing Decades of Buried Code Debt
Britain’s NCSC is sounding the alarm that AI-powered bug hunting is about to surface decades of quietly ignored security flaws — the kind that were easier to bury than fix. The irony is rich: AI creates new attack surfaces while simultaneously exposing every shortcut developers took since the Clinton administration. IT and security teams, you might want to cancel your summer plans.
AI-Generated Actors and Scripts Are Now Ineligible for Oscars
The Academy has drawn a hard line: fully AI-generated performances and scripts won’t be considered for Academy Awards. It’s a principled stand, though enforcing “fully AI-generated” in a world where every writer uses AI assistants and every VFX house runs AI tools is going to require some very creative rule interpretations. I give it three award cycles before there’s a controversy that makes everyone wish they’d defined their terms more carefully.
Microsoft Wants Lawyers to Trust Its New AI Agent in Word Documents
Microsoft’s new Legal Agent in Word promises to handle contract review, negotiation history, and complex document management using “structured workflows shaped by real legal practice” — which is Microsoft’s way of saying they know lawyers won’t touch something that just wings it. Convincing a profession built on adversarial skepticism to trust AI with billable work is either a masterstroke or a very long sales cycle. My money’s on the latter.
Mythos Complicates the Breakup, Says Pentagon CTO, but Anthropic Is Still Barred
The Pentagon’s CTO is clarifying that government agencies evaluating Anthropic’s cybersecurity-focused Mythos model doesn’t mean the DoD and Anthropic are reconciling — they’re just… window shopping. It’s a very Washington way of saying “we’re interested but not committed,” and it tells you everything about how the national security community is trying to thread the needle between AI capability and institutional trust issues with specific vendors.
Bottom Line
The week’s throughline is accountability — in courtrooms, in award ceremonies, at theme park gates, and buried in twenty years of unpatched code — and AI is forcing all of it into the open at once.