Senator Van Hollen Presses OpenAI’s Altman Over GPT-6 Astra Safety

This marks a shift from voluntary AI safety pledges toward formal congressional pressure, and it lands specifically on the national-security and critical-infrastructure angle rather than general model ethics. The naming of the Hugging Face incident signals that lawmakers are now treating prior containment failures as precedent, not one-offs. The near-term signal is whether OpenAI answers substantively by the September 17 deadline or defaults to a public statement, and whether Van Hollen's request for federal researcher access becomes a template other senators copy.