Sam Altman told a federal court this week that he is “an honest and trustworthy businessperson.” He said this under oath, on the witness stand, in a case where the central question is whether he and the organization he runs broke faith with its founding mission. It was that kind of week for AI in court.

The Trial Nobody Saw Coming, Now Impossible to Look Away From

Musk v. Altman, filed in the U.S. District Court for the Northern District of California, is probably the most consequential AI governance case being litigated right now, even if the week’s headlines made it sound like an episode of something on HBO.

The case goes back to OpenAI’s origins. When Elon Musk co-founded OpenAI in 2015 and contributed substantial early funding, the organization was structured as a nonprofit with a stated mission to develop AI for the benefit of humanity broadly. Musk’s core legal argument is that OpenAI’s conversion to a commercial structure violated that mission and the commitments that came with it, and that the assets built up under charitable auspices can’t legally be handed over to private investors under California nonprofit law. He’s asking for an injunction and damages.

OpenAI’s counter is sharper than you’d expect: that Musk knew the organization would need capital to compete at the frontier, that he was offered a central leadership role he walked away from, and that this lawsuit is a competitor using the courts to harass a rival. They’ve been pretty direct about saying that last part.

What made this week different is that both sides put on evidence in trial, and that’s where things got theatrical. Wired reported that OpenAI introduced a physical trophy as evidence, claiming it was proof of Musk’s “concerning behavior.” The item was, apparently, an ass. Whether it moves the needle legally is a different question, but it’s now in the federal record.

Altman’s testimony about his own honesty is the more substantively important moment, even if it sounds like a punchline. His credibility matters because the case involves contested accounts of what he told board members, investors, and co-founders over a decade of organizational change. “I believe I am an honest and trustworthy businessperson” is the kind of statement that opposing counsel writes down and uses in closing.

Why This Case Actually Matters

Strip away the tabloid texture and there’s a real governance question at the center of Musk v. Altman, one that the AI field hasn’t had to seriously grapple with before.

OpenAI was the template for a particular theory: that you could build potentially dangerous technology responsibly if you structured the organization as a nonprofit, insulated from shareholder pressure. A lot of people, including some major donors, gave money on that basis. The conversion to a for-profit structure raised roughly $157 billion in valuation and gave investors real equity stakes. Whether that required regulatory approval under California Attorney General supervision is still being sorted out separately from this trial.

What Musk is testing, effectively, is whether there are any legal teeth behind the charitable mission framing that early AI safety organizations used to attract support. If the court decides OpenAI can convert freely, it signals that nonprofit AI organizations can change their structure whenever it becomes inconvenient. If Musk wins on some version of his claims, it puts every similar organization on notice that their mission documents have actual legal weight.

Neither outcome is clean. But the case is forcing the question, and that has value even if you think Musk’s motives are entirely competitive.

Meanwhile, xAI Has a Different Kind of Legal Problem

While Musk’s lawyers are in California arguing about OpenAI’s obligations, his own AI company is facing a lawsuit over how it’s powering its Colossus 2 data center.

According to TechCrunch, xAI has been running nearly 50 gas turbines at the facility, and the legal problem is how those turbines are classified. Under the Clean Air Act, stationary sources of air pollution, meaning permanent industrial power plants, require permits that carry serious emissions review and limits. Mobile or temporary generators occupy a different regulatory category with lighter requirements.

The lawsuit alleges xAI has been running what are functionally permanent power plants while classifying them as “mobile” units to avoid the permitting process. The distinction matters enormously in terms of what the company had to disclose about emissions, what neighbors were told, and what mitigation obligations apply.

This is not a niche technical dispute. Data centers for large AI training runs are extraordinary power consumers, and the infrastructure buildout to support them has run well ahead of utility grid capacity in most markets. xAI’s workaround, if the allegations are accurate, reflects a broader problem: companies that need to move fast on AI infrastructure are running into the reality that environmental permitting processes weren’t designed for this pace. The temptation to find regulatory arbitrage in equipment classification is real and, apparently, legally risky.

Who filed the suit and in which court wasn’t included in the initial reporting, but the case will be worth tracking as more AI data centers face similar scrutiny from neighbors and regulators.

Also Worth Watching: Researchers Getting Squeezed Out

On the regulatory front, a less-covered case got a hearing Wednesday that has real stakes for AI policy research.

In Coalition for Independent Technology Research v. Rubio, filed against Secretary of State Marco Rubio in U.S. District Court, the nonprofit CITR is challenging a Trump administration policy that allows restricting visas to foreign nationals who advocate for content moderation policies at American tech platforms. Judge James Boasberg heard oral arguments in the case.

The AI connection is direct: content moderation and AI governance research increasingly overlap, and a significant portion of the academics working in this space are foreign nationals or based abroad. A policy that can be used to keep critics of American tech platforms out of the country has obvious chilling effects on the kind of international research collaboration that informs regulation.

The administration’s position is that it has broad discretion over visa decisions and that advocacy for foreign-style content rules is a legitimate basis for restriction. CITR argues the policy effectively punishes researchers for their speech and academic work, raising First Amendment concerns that don’t disappear just because the target is a non-citizen.

Boasberg, who has ruled against the administration in other high-profile cases, will decide whether the plaintiffs have standing to challenge the policy and whether it survives constitutional scrutiny. The outcome will determine whether foreign AI and tech policy scholars can safely attend American conferences, testify before Congress, and collaborate with U.S. institutions without worrying that their published views create visa problems.

Three courtrooms, three different theories of accountability. What they have in common is that someone decided AI companies and the people running them needed to answer for something. That’s new, and it matters.