Anthropic said on February 12, 2026 that it is contributing $20 million to Public First Action, a new bipartisan 501(c)(4) organization that promotes public education about AI and policies such as frontier model transparency, a federal governance framework and export controls on AI chips. The donation moved one of the leading frontier AI labs directly into the political contest over how AI should be regulated in the United States. Anthropic added a second $20 million in July 2026, bringing its support to $40 million. For developers, the gifts are a reminder that the rules governing AI products in the US remain contested and could change quickly.
Key Facts#
- Donation: On February 12, 2026, Anthropic announced a $20 million contribution to Public First Action.
- Recipient: Anthropic describes Public First Action as a new bipartisan 501(c)(4) founded and led by Republican and Democratic strategists.
- Stated priorities: Transparency safeguards for frontier AI companies; a robust federal AI governance framework, with opposition to preempting state laws unless Congress enacts stronger safeguards; export controls on AI chips; and targeted regulation of AI-enabled biological weapons and cyberattacks.
- Public opinion: Anthropic cited recent polling in which 69% of Americans said the government is "not doing enough to regulate the use of AI."
- Second gift: On July 21, 2026, Anthropic contributed another $20 million, bringing its total to $40 million.
- Limits: Anthropic says both donations support only Public First Action's public education and policy mission and cannot be used to influence the election of any candidate for federal, state or local office.
What Happened#
Anthropic's announcement opened with the argument that AI's benefits come with serious risks. It pointed to AI already being used to automate cyberattacks, as in the AI-orchestrated espionage campaign it disrupted in 2025, to the possibility that future models could help produce dangerous weapons, and to the risk of powerful systems taking harmful actions their users never intended. To illustrate the pace of progress, the company said it had redesigned a notoriously hard technical test for hiring software engineers several times because successive models kept beating it.
From there, Anthropic argued that AI policy decisions over the next few years will touch nearly every part of public life, from the labor market and online child protection to national security, and that it does not want to sit on the sidelines. It said there are no official guardrails in place and no federal framework on the horizon, that few organized efforts exist to mobilize people and politicians who understand the stakes, and that "vast resources" have flowed to political organizations that oppose such efforts. It did not name those organizations.
The company stressed that the policies it supports are neither partisan nor self-serving. Effective AI governance, it wrote, means more scrutiny of companies like Anthropic, not less. It added that transparency regulation should apply only to companies developing the most powerful and most dangerous models, so it would not hold back smaller or less well-resourced developers.
Background#
The donation built on positions Anthropic had argued for more than a year. In July 2025 it proposed a transparency framework for the largest frontier developers. In September 2025 it endorsed California's SB 53, which Anthropic later described as establishing the nation's first frontier AI safety and transparency requirements for catastrophic risks. It also opposed a proposed 10-year moratorium on state AI laws, while saying it would prefer one national standard to a patchwork of state rules.
Preemption is the thread that connects these positions. Anthropic's stated view is that Congress should not override state AI laws unless it puts stronger federal safeguards in their place. Public First Action's agenda, as Anthropic described it, sets the same condition, which aligns the group closely with the transparency-first approach Anthropic had promoted in California and in Washington.
Why It Matters for Developers#
- Expect regulatory churn. Political money on more than one side means AI rules will be contested in state legislatures, in Congress and in public campaigns. Treat compliance as configuration: keep AI features behind flags, record which model serves which jurisdiction and centralize policy checks, as outlined in the Responsible AI and LLM security guide.
- More disclosure is coming from providers. The transparency measures Public First Action backs would put more safety frameworks, risk reports and evaluation results in the public domain. Build a process to use them in vendor reviews; the productionizing AI interview guide covers how teams evaluate and replace models.
- Vendor policy is a dependency. An AI provider's policy positions can affect its products and its customers, as the Pentagon supply chain risk dispute showed later in February. Keep switching costs low by calling models through
IChatClientfrom Microsoft.Extensions.AI, with evaluation suites ready for alternatives.
What's Next#
Anthropic's second $20 million, announced on July 21, 2026, came with a sharper argument. The company pointed to Claude Mythos Preview, which it said had found thousands of high-severity software vulnerabilities, including in every major operating system and browser, and which it released only to a limited set of cyber defenders through Project Glasswing. It said transparency alone is now insufficient and pointed to its Advanced AI Framework, under which governments could verify companies' safety claims, enforce safe practices through civil penalties and slow or block the deployment of models that pose a serious risk of catastrophic harm.
The framework would apply only to models trained with more than 10^25 FLOPs by companies with more than $500 million in AI-related revenue or more than $1 billion in annual AI R&D spending. It says Congress should not preempt state law unless it enacts a federal law at least as strong as the framework. Anthropic also backed tighter export controls on advanced chips and chipmaking equipment and efforts to curb illicit model access and distillation attacks.
Whether the spending changes any votes cannot be judged from Anthropic's announcements. The clearer signal is that the debate over US AI regulation now involves significant political money, including, by Anthropic's account, from organizations that oppose the safeguards it supports.