On July 21, 2026, Anthropic announced it will give another $20 million to Public First Action, bringing its total contributions to the organization to $40 million. The company said both donations are restricted to supporting Public First Action’s public education and policy work on AI and may not be used to influence the election of any federal, state, or local candidate.
Reason for the increased support
Anthropic pointed to the rapid advance of AI capabilities as the main reason for the expanded funding. Earlier in 2026, the Claude Mythos Preview model discovered thousands of high-severity software vulnerabilities, including some affecting every major operating system and browser. Anthropic limited the model’s release to a set of trusted cyber defenders via Project Glasswing so those actors could find and fix weaknesses before they could be exploited.
The company notes that more capable models can deliver substantial benefits—speeding drug discovery, enabling treatments for previously untreatable diseases, accelerating scientific progress, extending human lifespans, and driving broad economic growth. However, Anthropic argues these benefits can only be realized if risks are addressed first, and governments need time to build capacity to capture benefits while containing harms. That urgency is a principal rationale for immediate policy engagement and public education.
Policy positions and proposals
Anthropic has long advocated for transparency from frontier AI developers about what their models can do and how they manage risks, and the company has supported recently passed state laws that require greater transparency. Yet Anthropic contends transparency alone is insufficient given the speed of capability improvements.
The company reiterated elements of Anthropic's Advanced AI Framework as the core policy response. Key points include enabling governments to verify companies’ safety claims, enforcing safe practices through civil penalties, and retaining the ability to slow or block deployments of models that pose a serious risk of catastrophic harm. Frontier AI developers should be required to test models that could pose catastrophic risks, publish findings for public scrutiny, submit models to independent evaluation, and maintain robust security programs.
Anthropic also supports policy measures to tighten export controls on advanced chips and semiconductor manufacturing equipment and to restrict illicit model access and distillation attacks, arguing these steps help prevent democracies’ best technologies from being used to advance authoritarian AI.
Next steps
Anthropic framed the policy frameworks it has proposed as starting points rather than finished solutions. The additional donation to Public First Action is presented as a way to raise the profile of the policy debate and public education around AI safety and governance. The announcement did not specify further financial commitments or programs on the part of Public First Action.
Notable names referenced: Anthropic, Public First Action, Claude Mythos Preview, Project Glasswing, Anthropic's Advanced AI Framework, 2028: Two Scenarios for Global AI Leadership.
