Americans for Responsible Innovation proposes AI firms disclose self-use of AI in model development

Nextgov/FCW reports, in an October 26, 2026 article, on a proposal from Americans for Responsible Innovation (ARI), shared first with the outlet, that would require AI companies to disclose the extent to which their own systems help build new models. The aim is to help Washington track AI advances and judge whether developers test safety adequately. ARI, a group pushing for federal AI oversight, wants government to to know how much research AI is doing and how closely people supervise it.

Under the plan, major developers would file quarterly reports with a newly created Commerce Department office, and report significant research changes inside four business days. The short notice could be triggered by something like automating an entire research cycle. Commerce could ask follow-up questions, require reports more often, and pass filings it deems knowingly false or incomplete to the Justice Department. A whistleblower program is also proposed, along with standards covering hacking capabilities, mass influence campaigns, and CBRN-type and explosive threats. It would use technical analysis from an existing NIST-based center in Commerce to shape safety standards and decide which models fall under federal oversight, and it leaves open whether responsibility would sit in Commerce or be shared among agencies.

ARI argues that automating model development could speed progress and leave officials less time to assess risks, and that what the government can see should not hinge on voluntary company announcements. The article places the proposal against a White House that is deepening its involvement with AI firms while resisting a broad new regulatory regime. President Trump recently put Jay Clayton, the Director of National Intelligence, in charge of a task force set up by executive order, which has assigned a Pentagon official to draft breach-reporting protocols.