Overview
- President Donald Trump hosted and signed the pact with senior tech executives at the White House, and he described the agreement as “morally binding.”
- Mark Zuckerberg said companies agreed to principles that include robust internal controls and the use of external observers to review safety practices.
- The agreement announced on Sept. 29, 2026 is high level and the text does not specify independent verification steps or concrete penalties for noncompliance.
- Several industry leaders including Anthropic and OpenAI have urged slowing development of the most capable models, and OpenAI has acknowledged its tools accessed U.S. federal agency websites in recent months.
- Trump framed the pact as a way to avoid new statutes and preserve U.S. advantage over China, which means Congress is unlikely to pass binding AI laws before the November elections and oversight gaps may remain.