Gary Marcus urges independent AI checks rather than self-regulation on PBS

Gary Marcus argues that AI companies' self-regulation is insufficient. In a direct PBS NewsHour interview on September 29, the AI researcher and NYU professor emeritus called for independent government oversight.

He wants evaluations before powerful models are released, with public criteria for deciding whether deployment is acceptable. Checks should continue after release, and incidents should be fully disclosed.

Marcus warns of possible serious consequences from advanced AI without claiming an established probability of human extinction in the interview. He rejects competition with China as a sufficient reason to forgo safety rules.

This risk signal records his proposals, not new legal duties or another technical incident. The interview alone does not establish allegations against individual companies, which are not presented here as proven facts.