Mark Zuckerberg rejected a coordinated AI-development slowdown in his September 24, 2026 NBC interview. Individual labs could withhold unsafe products themselves.
He points to customer trust as an incentive for safer agents. Whether that incentive provides sufficient control remains his judgment.
In his September 16 original post, Zuckerberg had already said Meta delayed Muse for safety work. He also supported independent evaluators.
The interview develops that counterposition; it establishes no industry consensus. Meta's safety claims are not independently validated by these statements.