Musk v. Altman: when asked whether xAI has distilled OpenAI models, Elon Musk says the claim is “partly” true
While answering questions under oath, Musk argued it's standard practice for AI labs to use their competitors' models. — While testifying on Thursday …
Wired
Context & Ripple Effects
The related coverage tracks a dispute that broadened from Musk’s own testimony to competing accounts of OpenAI’s leadership and safety practices, then to closing arguments centered on credibility and evidence. This testimony matters because it puts xAI’s claimed use of a rival’s model directly into the record rather than leaving it as an allegation.
The corpus also identifies xAI as a lab expanding its compute footprint and pursuing deployments of Grok, making the handling of model-access practices consequential for a company seeking to scale its own systems.
First-order effects
Musk’s partial acknowledgment gives OpenAI and opposing counsel a clearer factual basis to probe what xAI used, how it was used, and whether those practices crossed any contractual or legal boundary.
xAI must defend the distinction between industry-standard model use and the specific conduct at issue, while its public claims about independently developed frontier systems face closer scrutiny.
Second-order effects
The dispute raises the value of auditable controls around API access, output collection, and training-data provenance for AI labs that expose models to competitors and third parties.
Rival labs may tighten terms, monitoring, and technical limits on model access if litigation makes ordinary evaluation or output use harder to distinguish from distillation.
Third-order effects
If courts and labs increasingly treat model-output reuse as a contested competitive input, frontier-model competition could shift toward more controlled access regimes and stronger provenance requirements rather than open-ended API experimentation.
The case illustrates that legal definitions of permissible model-to-model learning may become a meaningful constraint on AI competition, though the corpus does not establish how the dispute will be resolved.
The trend: This is one data point in the shift from treating model outputs as a broadly usable product surface to treating them as a governed and litigable source of competitive advantage.
Musk doubled down that everyone is doing this. He also said it's standard practice for AI labs to use other AI models to validate their own, which seemed more known. https://www.wired.com/...
Elon's lawsuit against OpenAI is really not about Grok. But a line of questioning this morning veered into how xAI trains its AI models, potentially using their competitor's technology: Lawyer: Do you know what distillation is? Musk: It means to use one AI model to train another …
What people call “distillation” is a super common practice (you use other models to benchmark your model, to evaluate your inputs or to add a little bit to your datasets) that in my opinion should be covered by fair use (just like using public data is), especially when the
I think OpenAI did basically betray its commitments to its early investors but it's so absurd to have Musk being the one suing when he 1) didn't seem to care until it affected him personally and 2) is now running a fully for-profit competitor.
LIVE TRIAL UPDATE: OpenAI's counsel asked Musk whether xAI has ever “distilled” technology from OpenAI. Musk: “Generally AI companies distill other AI companies.” “Is that a yes?” Savitt asked. Musk: “Partly.”
it is quite significant that Musk admitted on the stand that xAI is distilling OpenAI models to train xAI, and that it is using OpenAI's technology to build xAI!
lmao — OpenAI's lawyer piped up and both him and Musk's counsel started raising their voices until the judge said they all need to shut up nothing like being dressed down as an adult
judge Gonzalez: “I suspect that there are a number of people who do not want to put the future of humanity in Mr. Musk's hands. But we're not going to get into that. This is not a trial on the safety risks of artificial intelligence.”
i am not really sure how often lawyers try to endear themselves to judges but Musk's lawyer, Steven Molo, does not seem to be trying to do that right now he's trying to get “extinction risk” discussion into the court discussion. “This is a real risk. we all could die.”
Musk also told a Neuralink exec they could hire directly from OpenAI, and pitch people who work there to come to Neuralink. On the witness stand, Musk tried to argue it would be illegal for his companies not to hire people from OpenAI. Musk also said Andrej wanted to leave [image…
New: Musk was cross-examined by OpenAI's lawyers in federal court today. They pressed him on all the ways he squeezed OpenAI over a power struggle in 2017, including by hiring away talent and cutting off funding. Here's an email Musk sent about hiring Andrej Karpathy. [image]
Musk had also been sending $5 million quarterly payments to OpenAI as part of a larger $1 billion funding commitment he made at the org's launch. Around this time, he instructed the head of his family office, Jared Birchall, to stop paying them. Musk said this was because he [ima…
The thing about suing someone is they can drag up old emails you sent and read them out loud in front of everybody which tbh sounds like a GREAT reason never to sue anybody ever at all. from @mzeff.bsky.social
Judge cont: “I suspect there are plenty of people who dont want to out the future of humanity in Mr. Musk's hands, but it doesn't matter, we aren't going to get into those issues. We're just not going to do it, and ultimately, it's an advisory opinion. ....This is not a trial on
Intense argument right now between the attorneys (no jury yet) about whether an expert will discuss AI safety risks “WE ALL COULD DIE” Musk's attorney says. Judge isn't having it: “I said stop! ... It is also ironic that your client, despite these risks is creating a company that
Last week: White House condemns Chinese AI labs for alleged distillation attack This week: Elon admits under oath that xAI distilled OpenAI models to train its models 🫣 [image]