Overview
- The administration signed the executive order Tuesday, June 2, 2026, asking leading AI labs to voluntarily give the government access to their most powerful models up to 30 days before public release.
- The order directs the Treasury secretary to set up a cyber information‑sharing center for AI to detect, validate and coordinate fixes for software vulnerabilities affecting critical infrastructure.
- The policy was shaped by recent doubts about Anthropic’s Mythos model and by internal debate that briefly produced a draft with a 90‑day review window before advisers helped narrow it to 30 days.
- Officials say the measure is voluntary and does not create a new mandatory licensing or pre‑authorization regime, but implementation rules, which firms will participate, and review procedures remain to be defined.
- The move aims to reduce AI‑driven cyber threats and protect infrastructure while keeping U.S. firms competitive, and it follows a wider global push by governments to oversee advanced AI without halting commercial development.