A trial has been underway in a federal courthouse in Oakland this month, and most of the coverage has framed it the same way: Elon Musk is suing OpenAI's leadership, alleging they betrayed the company's founding mission as a public charity when they restructured it into a for-profit operation. Sam Altman and Greg Brockman have testified. Shivon Zilis, who served on OpenAI's board, has testified. More executives may still be called.
The legal question is whether Altman and Brockman violated a charitable trust. The verdict will be advisory only — the judge has the final say on the legal claims. Either way, plenty of people will read the outcome through whatever lens they brought into it. Musk-aligned readers will see a vindication or a defeat. OpenAI-aligned readers will see the opposite. The thing they'll mostly miss is what the trial is actually about.
I've been watching the testimony, not the headlines, and what's emerging is bigger than either of the parties.
The thing the headlines aren't capturing
The trial is producing a long, on-the-record account of how internal governance has worked at one of the most important companies of our era. And the picture that's emerging is unflattering — not in a partisan way, but in a structural way.
Rosie Campbell, a former member of OpenAI's AGI readiness team, testified that her team was disbanded in 2024. The company's Super Alignment team was shut down in the same period. Campbell described joining a research-focused organization where AGI and safety were common topics of conversation, and watching that orientation gradually disappear as the for-profit subsidiary expanded. Tasha McCauley, a former board member, described a pattern of internal information-control that made it difficult for the non-profit board to exercise oversight of the for-profit operation underneath it.
You can hold a range of views on the merits of Musk's case and still find this picture concerning. The witnesses describing the internal dynamics are former employees and former board members who were inside the building. They're not partisans; they're the people the public was relying on to keep an eye on the most powerful AI systems being built. Their testimony, taken at face value, suggests that the internal checks on the most powerful AI company in the world have been weaker than the public has assumed.
McCauley said something in her testimony that's stuck with me. She said the failures of internal governance at OpenAI should be a reason to embrace stronger government regulation — because "if it all comes down to one CEO making those decisions, and we have the public good at stake, that's very suboptimal."
That's not a partisan claim. That's a structural one.
Why this affects everyone, not just AI insiders
When the safety teams inside an AI lab are disbanded, that decision shapes what gets built. When the board can't reliably get accurate information from the CEO, that shapes which products ship. When governance is weak, the products that emerge are products optimized for whatever the company optimizes for — and right now, what the major AI companies optimize for is engagement and revenue.
The design pattern at the heart of this is what produces the AI that always agrees with you. The AI that mirrors your worldview. The AI that's optimized to keep you in the app. None of those are random outcomes. They're what happens when the safety teams are smaller than the growth teams.
I'm not saying the trial will fix any of this. Trials don't fix industries. The trial will produce a verdict, OpenAI will continue to operate, and the next quarterly product cycle will arrive on schedule. The structural conditions that produced the testimony will mostly stay the same.
What the trial does do is make the structural conditions visible. Public testimony, on the record, from people who were there. The picture is now part of the public record in a way it wasn't six months ago.
The other lesson — the one for the rest of us
There's a version of this story where the conclusion is "regulation will save us." I don't disagree with the regulation argument, but I want to point at something else, because regulation moves slowly and AI moves quickly.
The other lesson — the one available right now, today — is that the public bought a story about the AI industry that the industry can't actually deliver on. The story was: trust us, we have safety teams, we have boards, we have alignment researchers, we're being responsible. The trial is making the gap between that story and the reality very visible.
Once you see the gap, the question changes. It's not "how do we make these companies more trustworthy." It's "what are we willing to put into a system whose internal trustworthiness we can't verify?"
The honest answer for most personal use is: less than what we've been putting in.
The thinking you do for yourself. The questions you'd never ask out loud. The decisions you're working through. The reflections you'd want shielded if anyone came asking. None of that needs to live on a server controlled by a company whose internal governance is currently being litigated in federal court.
What we built instead
The reason Blob exists is not that we predicted any particular trial. It's that we noticed, several years ago, that the architecture of the major AI products forced the user to trust the company's promises about what they'd do with your data. And we noticed that companies, in the long run, do whatever the structural pressure on them forces them to do — not whatever they originally promised.
So we built something where the trust isn't required. The architecture itself is the protection. Your conversations are encrypted end-to-end. We can't read them. We can't sell them. We can't be ordered to produce them, because they don't exist on our servers in a form that matches what you typed.
See what an AI on your side actually feels like. Free to try. No credit card.
The trial will end. The verdict will arrive. The headlines will move on. The structural conditions that produced the testimony will mostly stay the same.
What you put into the system is up to you.