Trump's reported AI tsar pick would hand AI policy to the nation's intelligence chief

On paper, Clayton has an enforcer's resume. He chaired the Securities and Exchange Commission from 2017 to 2020 ^[5], served as U.S. Attorney for the Southern District of New York, and was earlier a partner at Sullivan & Cromwell ^[6]. On AI, though, he talks about security. He has opposed calls to pause the development of AI models ^[11]. "Superintelligence is a national security issue," Clayton said in a CNBC interview the day after a White House meeting with technology executives. "That was something that was recognized yesterday by not just the industry leaders, but everybody in the room." ^[15]

His boss is lighter still. The administration has favored U.S. dominance in the field over heavy regulatory frameworks ^[13]. "I think I'm seeing tremendous self-policing," Trump told reporters at the White House. "And when they're not, we're going to be able to nab them." ^[7] An official who opposes development pauses, working for a president who expects the industry to police itself, is a weak basis for forecasting a general crackdown on the companies investors are funding. Clayton did say the administration would take a "whole-of-government approach" to regulating the technology ^[16].

There are three plausible versions of the job. It can stay a title with convening power. Its duties are undefined ^[8], it would come on top of running the intelligence community ^[24], and Al Jazeera reported it was unclear whether Clayton would hold both posts ^[25]. It can become a security office that acts on access; Crypto Briefing noted that security-driven policy "tends to emphasize control over who builds what, who gets access, and where technology flows" ^[26]. Or the "AI Force" agency Trump has floated could take the operational work, though how a tsar would fit beside it has not been spelled out, according to Crypto Briefing ^[9].

The second version already has a test case, and it is a company preparing to list. The appeals court that cleared the Pentagon blacklist found the agency "had ample support for its conclusion" that Anthropic's products are a supply chain risk because of code that prevents them from performing certain tasks ^[18]. Anthropic has said it will not allow its technology to be used for autonomous weapons or surveillance ^[19]. The government risk in its prospectus ^[20] follows from a choice written into its own product, and the offering is aimed for next month ^[21].

The staffing choice shows how much of one official the administration is assigning to AI. Sacks, a venture capitalist ^[3], held a dedicated title, special adviser for AI and Crypto, for about 15 months to March ^{[4] [22]}. The seat has been empty for about six months since ^[23]. Clayton would take it as a second job ^[24]. The administration is not appointing a full-time AI official, and the AI file would compete with oversight of the spy agencies for one person's hours.

In my view, the appointment raises the cost of refusing defense and intelligence customers and leaves the rest of the sector under the self-policing line. That view would be wrong if the formal announcement gives the tsar rulemaking or enforcement power over model developers in general, or if Clayton starts asking companies with no government business for SEC-style disclosures. The appointment itself is unconfirmed: the reporting drew on anonymous sources, and Clayton has not said he wants the job ^[14].