All corrections
LessWrong March 6, 2026 at 05:16 AM

www.lesswrong.com/posts/rWc6BT8BcoPhwcbSs/ai-safety-has-12-months-left

1 correction found

1
Claim
the US (which revoked its own executive order and used the DPA against the company that held the line)
Correction

Reporting on the Anthropic–Pentagon dispute described the Defense Production Act (DPA) as a threat/potential escalation, not something the U.S. actually invoked or "used" against Anthropic.

Full reasoning

The post asserts that the U.S. "used the DPA" against Anthropic (the “company that held the line”).

However, contemporaneous reporting about the February 2026 Anthropic–Pentagon dispute states the Pentagon threatened it could invoke the Defense Production Act, and that Trump’s directive stopped short of those threats.

  • Al Jazeera (AP/Reuters) (Feb 27, 2026) reports that Trump’s decision "stopped short" of Pentagon threats, explicitly including the possibility of invoking the DPA—i.e., the DPA was described as something that could be invoked, not something that was invoked.
  • Mayer Brown (Mar 2, 2026) likewise describes DPA invocation as something Hegseth "had threatened" earlier, and notes the administration had not publicly identified the specific legal authority it would invoke—indicating the public record did not show the DPA being used.

Given these sources, the post’s phrasing that the U.S. "used the DPA" overstates what is publicly evidenced (threatened/potential vs. actually invoked).

2 sources
Model: OPENAI_GPT_5 Prompt: v1.6.0