What happened
On June 10, 2026, Anthropic released a paired policy package comprising two substantive framework documents alongside an accompanying essay by CEO Dario Amodei. The Advanced AI Framework (AAIF) proposes mandatory obligations for developers of models trained using more than 10²⁵ FLOPs, or companies earning over $500M in AI revenue or spending more than $1B on R&D — covering transparency, mandatory third-party evaluation, security requirements for model weights, and government authority to block or deter deployments posing biological, cyber, loss-of-control, or automated R&D risks. Anthropic explicitly compares the envisioned regime to the FAA: "Frontier AI models, like airplanes, should be required to go through technical testing and auditing, and their release should be blocked or reversed as a threat to public safety." The Economic Policy Framework (EPF) acknowledges that AI "may become a general substitute for human labor" and proposes a tiered US policy response to labour displacement, backed by $350M in committed funding ($200M for an Economic Futures Research Fund, $150M for a national fellowship programme). Amodei writes that the company's previous advocacy for transparency alone "is no longer sufficient" given evidence from Claude Mythos Preview's real-world cybersecurity impacts.
Why it matters
This is the first time a frontier AI lab has simultaneously proposed specific legislative thresholds (FLOPs, revenue) for mandatory oversight, called for government authority to block model deployments, and backed economic displacement policy with committed capital — raising the floor for what other labs and regulators must respond to.
Action needed
Brief your government-affairs and legal teams on the AAIF's specific thresholds; if your enterprise licenses frontier models, assess whether regulatory deployment holds or model revocations should be incorporated into your AI vendor risk scenarios.