Tech Policy Press commentary urges public inclusion in SAFA frontier-AI body governance

A commentary on Tech Policy Press argues that the planned Standards Authority for Frontier AI, or SAFA, should be built with democratic safeguards. It says OpenAI, Anthropic and Google are creating the lab-funded body to coordinate pre-deployment testing and incident reporting, modeled on the US financial industry's FINRA, which works only because the SEC oversees it. The author says SAFA lacks such a backstop and is unlikely to get one.

The piece sets this against recent US events: an FTC probe into AI harms to consumers, threats from President Trump toward companies that might slow development, and a voluntary agreement that company leaders signed at the White House, described as morally binding. The author reads the accord as a possible soft antitrust waiver, since it approves regular safety meetings among companies, while the administration keeps the threat of enforcement in reserve. Polling cited in the article shows 57% of Americans want immediate federal regulation and 86% would back independent safety standards even if development slowed.

The author defines legitimate governance as representative, transparent and accountable. For SAFA, that could mean voting seats for academics, independent experts and civil society, public comment on draft standards, mandatory public reporting of security incidents and audit results, and periodic review of the structure. Accountability would have to come from the charter, for instance independent audits, whistleblower protection and financial penalties or expulsion for persistent violators.

The article notes that SAFA is expected to launch as early as 2027, and that its charter, structure and leadership are not public as far as is known. That leaves a window, it argues, for the companies to involve outside voices now, in line with their own stated support for democratic governance.