Last Week in AI #340 - OpenAI vs Musk + Microsoft, DeepSeek v4, Vision Banana

The Musk v. Altman trial kicked off last week, and the opening days belonged almost entirely to Elon Musk on the stand. His core argument hasn't changed since he first sued: OpenAI was founded as a nonprofit meant to benefit humanity, and converting it into a for-profit vehicle amounts to, in his words, stealing a charity. That line landed hard enough that it's now the phrase attached to the whole case.

What makes this round more interesting than the usual Musk-Altman sniping on X is the paper trail. Court filings surfaced texts showing Musk reached out to OpenAI co-founder Greg Brockman about a settlement just two days before the trial began, which is an odd move for someone who's spent years publicly insisting OpenAI's leadership acted in bad faith. Either he wanted to avoid a costly, drawn-out legal fight, or the discovery process turned up something he didn't love having read into the record. Both are plausible.

Things got messier for Musk personally when he appeared to admit, under questioning, that his own company xAI had used OpenAI's models to help train Grok. If that holds up, it's a genuinely awkward look: a man suing a company for allegedly betraying its founding mission while his competing venture leaned on that company's technology to bootstrap its own. OpenAI's lawyers will almost certainly lean on that admission going forward, since it undercuts the moral high ground Musk has tried to occupy this entire saga.

Proceedings are now public on YouTube, so anyone curious can watch the depositions and cross-examinations directly rather than relying on secondhand courtroom reporting. Given how much of AI's near-term direction has been shaped by decisions made inside OpenAI's boardroom, and by fights over who controls it, that transparency matters more than a typical corporate lawsuit would.