Anthropic Backs Trump AI Executive Order with 30-Day Pre-Release Review
U.S. President Donald J. Trump signed the executive order "Promoting Advanced Artificial Intelligence Innovation and Security" on June 2, 2026, aimed at advancing AI innovation and strengthening cybersecurity. The same day, Anthropic praised it on its official X account as "an important step in strengthening America's leadership in AI," signaling its intent to collaborate with the White House on implementation.
The order directs federal agencies to prioritize the cyber defense of National Security Systems and civilian federal information systems, and establishes a voluntary framework for cooperation with AI developers. At its core is a mechanism for "covered frontier models"—designated cutting-edge AI models—under which the government may access these models for cybersecurity review up to 30 days before public release. Notably, mandatory licensing, pre-clearance, and permitting requirements are explicitly rejected, reflecting a consistent, innovation-first stance that avoids regulatory burdens. Full executive orderFact Sheet
Designation of covered models works through a classification benchmarking process led by bodies such as the NSA, which select models with advanced cyber capabilities as "covered frontier models," with developers able to work with the government to determine whether a designation applies. Stated goals include scanning for cyber vulnerabilities, providing AI cyber defense tools for critical infrastructure such as local hospitals, community banks, and local utilities, and coordinating vulnerability patches. The order also calls for Binding Operational Directives issued by OMB/CISA, the creation of an AI cybersecurity clearinghouse, and expanded federal cyber workforce hiring, with various deadlines set at 30 and 60 days.
The move comes after the Trump administration, since taking office in January 2025, rolled back AI regulations from the Biden era and rolled out a series of innovation-focused policies, including a July 2025 AI Action Plan and a December 2025 order addressing conflicts with state law. A draft order considered around May 2026 had debated a 90-day government pre-access window, but it was delayed and revised over competitiveness concerns, ultimately settling on a 30-day period and a voluntary framework focused on cybersecurity. It was an outcome reached after the White House infighting reported by Axios. CFR analyzes the shortening to 30 days as a decision to balance innovation and safety while preserving AI competitiveness against China.
Anthropic's cooperative stance is noteworthy. The company previously faced a dispute over guardrails in the use of Claude with the Pentagon, leading the government to designate it a supply-chain risk and issue a stop-use directive. While Anthropic had earlier been noted for resistance to government access, this time it clearly expressed willingness to support implementation. Competitors such as OpenAI and Google may also engage with similar frameworks, suggesting an industry-wide aim to strengthen public-private cooperation. Anthropic's official post
Reactions on X, coming immediately after the announcement, largely focus on the order as a whole. Positive voices stand out, calling it "a strong signal for US AI leadership," saying "policy is catching up to the capability curve," and framing it as "a win that puts innovation first without regulatory burden." Some praise the provisions establishing the voluntary, non-compulsory nature, while others view Anthropic's shift in posture as "pragmatic." techedgedailyaipulseda1ly Industry groups including SIIA and the Business Roundtable issued statements welcoming the balance of "innovation and security" and the "voluntary framework." The order itself is centered on a policy framework and does not include product specifications such as pricing, benchmarks, or release timing, and detailed developer-level use-case assessments remain limited for now.