The U.S. AI Safety Institute (housed at NIST) announced formal Memoranda of Understanding with Anthropic and OpenAI that grant pre-release and post-release access to frontier models for collaborative safety research, evaluation, and risk mitigation. The agreements formalize government access to industry models for independent capability and safety assessment, with feedback coordination across U.S. and U.K. safety institutes. For financial institutions, this represents a regulatory infrastructure point: government now has explicit, ongoing access to frontier capabilities and internal model details, creating a structural precedent for safety-conditioned model deployment and potential future leverage over vendor development priorities.
320 other tracked developments touching the same claims.