Anthropic, the artificial intelligence safety company founded by former OpenAI researchers, has been actively briefing the Trump administration on its AI safety research even as the company pursues litigation against the U.S. government, according to Jack Clark, the company’s co-founder and chief policy officer. Clark made the disclosure during remarks at the Semafor World Economy summit this week, clarifying what initially appeared to be a contradictory stance toward federal engagement.
The revelation underscores a pragmatic division within Anthropic’s approach to government relations—separating its legal disputes with federal agencies from its broader policy advocacy and technical collaboration on artificial intelligence governance. Anthropic, which has positioned itself as a leader in AI safety through its development of Claude, a large language model designed with constitutional principles, has been engaged in substantive discussions with the Trump administration about the company’s research direction and safety protocols. Clark’s disclosure suggests the litigation does not preclude ongoing technical and policy dialogue at senior levels.
The distinction matters because it reflects how major technology companies increasingly navigate complex relationships with government—simultaneously contesting regulatory actions, compliance demands, or funding decisions while maintaining channels for technical expertise sharing and strategic input on policy matters. For Anthropic, which was founded on principles of responsible AI development, this dual engagement strategy allows the company to advocate for its preferred regulatory framework while fulfilling what it views as a civic obligation to brief policymakers on AI safety considerations. The Trump administration, which has signaled its intent to accelerate AI development while maintaining national security guardrails, appears receptive to direct technical briefings from leading AI companies.
Clark did not provide extensive details about the specific content of Anthropic’s briefings to Trump officials or identify which administration representatives participated in such sessions. However, his acknowledgment at a high-profile economic summit suggests these interactions are neither covert nor exceptional—they represent standard practice within the AI industry’s engagement with federal policymakers. Anthropic’s Claude has become increasingly prominent in enterprise and government contexts, making the company a de facto stakeholder in U.S. AI policy formation. The company has invested significantly in Washington lobbying and policy outreach, hiring former government officials and establishing relationships across both Democratic and Republican administrations.
The nature and scope of litigation between Anthropic and the U.S. government remains partially opaque to public scrutiny, though technology sector observers have tracked various disputes between AI companies and federal agencies over data access, regulatory compliance, and competitive concerns. Clark’s remarks suggest that whatever legal disagreements exist with specific agencies or over particular policies, they do not reflect a fundamental rupture in Anthropic’s willingness to engage constructively with the executive branch on matters of national interest. This compartmentalization—legal contestation on one track, policy collaboration on another—is common in regulated industries, where companies may challenge regulatory authority while simultaneously providing technical input to those same regulators.
The disclosure carries implications for how AI governance will develop during the Trump administration’s tenure. By maintaining active briefing relationships with companies like Anthropic, federal policymakers gain access to cutting-edge technical expertise and can calibrate their approach to AI regulation based on industry input. Conversely, companies like Anthropic secure opportunities to shape the regulatory environment before rules are finalized. This dynamic potentially favors larger, well-resourced AI companies with established Washington operations over smaller competitors or non-American firms seeking to influence U.S. policy. It also suggests that the Trump administration, despite its rhetoric about government efficiency and deregulation, recognizes the strategic importance of AI development and maintains sophisticated engagement with the sector.
Looking forward, observers should monitor whether Anthropic’s legal disputes with federal agencies are resolved or escalate, and whether such resolution correlates with the company’s policy influence in Washington. The precedent of ongoing government engagement during litigation may also shape how other technology companies calibrate their own strategies. Additionally, the substance of Anthropic’s briefings about AI safety—what specific concerns the company is raising, what regulatory approaches it is recommending—will likely influence the contours of any formal AI governance framework the Trump administration develops. The company’s ability to maintain this dual engagement strategy without apparent contradiction suggests that U.S. AI policy, at least in its formation phase, will be heavily shaped by the companies themselves rather than by independent regulatory bodies or external stakeholder input.
