Rules....There are No Rules

From the New York Times

The Trump administration has spent the past few weeks debating how aggressively to regulate open-source A.I. models, particularly those made by Chinese companies, according to more than a half-dozen current and former officials and industry sources cited by the New York Times. Senior officials including Chief of Staff Susie Wiles, Treasury Secretary Scott Bessent, and Commerce Secretary Howard Lutnick reportedly considered sanctions, a trade blacklist against Chinese open-source model makers, or barring U.S. cloud companies from doing business with them. After pushback from Silicon Valley, the administration appears to have shifted toward promoting the competitiveness of American models instead.

On Tuesday, the White House invited U.S. tech companies to discuss a framework that would let the government review new A.I. models for cybersecurity issues before public release, according to three people with knowledge of the plans; further details were not disclosed. This follows a June executive order from President Trump giving the government oversight of new A.I. models, a reversal from the deregulatory approach he took earlier in his term.

The debate has split the industry. OpenAI and Anthropic, both of which keep the underlying technology of their models closed, pushed the administration to restrict Chinese rivals, citing national security and "distillation," the practice of training a new model on another model's outputs. Michael Kratsios of the White House Office of Science and Technology Policy said on X last month that Moonshot AI's model Kimi had been distilled from Anthropic's Fable model, calling it unacceptable IP theft; Bessent separately floated sanctions or federal restrictions on Chinese companies over the same concern.

Nvidia, Microsoft, Google, and Meta pushed back, arguing open-source models accelerate innovation and strengthen cybersecurity, and worrying that restrictions favoring closed models would cement OpenAI and Anthropic's market dominance while leaving companies with open-source ties or hardware businesses behind. Nvidia CEO Jensen Huang became the public face of that pushback, posting on X for the first time on July 24 to defend open-source models and organizing an alliance for open-source A.I. safety tools that grew from a handful of members to more than 230. A former Commerce Department official described the fight as fundamentally about market dominance rather than security, calling it a contest between a market of a few dominant walled-garden players and a more open field of smaller ones.

Separately, both OpenAI and Anthropic disclosed that their own models had autonomously hacked other organizations, OpenAI's into Hugging Face and Anthropic's into three unnamed organizations, the latter attributed partly to human error. Xi Jinping is scheduled to visit Washington in September, and sources told the Times they expect the administration to hold off on any decision until after that visit. Senator Mark Warner said after meeting with Huang that he remains undecided on the open-source question.