This Is the Policy.

Anthropic’s News Policy feed lists 51 articles and, read in date order, the entries form an Anthropic AI safety policy timeline that moves from consumer terms and government alignment to independent evaluation, export-control suspension, and a data-center electricity commitment. The record begins on Aug 27, 2025 with the formation of the Anthropic National Security and Public Sector Advisory Council and ends on Oct 15, 2026 with employee-like access for a team from METR, a nonprofit that evaluates frontier AI models.

Building the Government and Consumer Baseline in 2025

On Aug 27, 2025, Anthropic announced the formation of the Anthropic National Security and Public Sector Advisory Council, a group of bipartisan national security and public policy practitioners meant to help the company support the US government and closely allied democracies in building and maintaining enduring technological advantages in an era of strategic competition.

The next day, Aug 28, 2025, Anthropic updated its consumer terms and privacy policy so Claude Free, Pro, and Max users could opt in to having conversations used to improve models, with a five-year retention period for participants. On Sep 4, 2025, the company said it was updating restrictions on sales to unsupported regions and strengthening regional restrictions to account for that reality and better align with its commitment to ensuring that transformative AI capabilities advance democratic interests. The same day, Anthropic signed the White House pledge to America's youth, investing in AI education for students and educators nationwide.

On Sep 8, 2025, Anthropic endorsed SB 53, the California bill that governs powerful AI systems built by frontier AI developers like Anthropic. On Oct 21, 2025, CEO Dario Amodei issued a statement on Anthropic’s commitment to advancing America's leadership in building powerful and beneficial AI. The year closed on Dec 19, 2025 with Anthropic’s Frontier Compliance Framework, which discloses how the company assesses and manages catastrophic AI risks under California's Transparency in Frontier AI Act.

Safety Documents, Energy Costs, and War Department Questions in Early 2026

On Jan 22, 2026, Anthropic published Claude's new constitution, described as a new approach to a foundational document that expresses and shapes who Claude is. On Feb 11, 2026, it committed to covering consumer electricity price increases caused by its data centers, including grid upgrades and new power generation.

On Feb 12, 2026, Anthropic donated $20 million to Public First Action, a 501(c)(4) focused on AI issues in the public interest. On Feb 24, 2026, the company released Anthropic’s Responsible Scaling Policy: Version 3.0, an update to its policy to mitigate catastrophic risks from AI. On Feb 26, 2026, CEO Dario Amodei issued a statement on Anthropic’s discussions with the Department of War and national security uses of AI. On Feb 27, 2026, Anthropic published a statement on comments from Secretary of War Pete Hegseth, including its response and advice for customers. On Mar 5, 2026, Amodei issued another statement on where things stand with the Department of War.

Research, Competition, and Export-Control Limits in Mid-2026

On May 7, 2026, Anthropic shared the focus areas for The Anthropic Institute, saying it would use the information it could access from within a frontier lab to investigate AI’s impact on the world and share its learnings with the public. The same post shared the questions that drive the institute’s research agenda.

On May 14, 2026, Anthropic published 2028: Two scenarios for global AI leadership, laying out its views on the AI competition between the US and China. On May 19, 2026, it said it had been organizing dialogues over the past several months with groups whose work and traditions bear on the questions raised by frontier AI.

On Jun 10, 2026, Anthropic released Policy on the AI exponential, saying AI is advancing at exponential speed while the policymaking process was built for a slower world, and sharing policy proposals to prepare institutions for AI progress. On Jun 12, 2026, Anthropic issued a statement on the US government directive to suspend all access to Fable 5 and Mythos 5.

On Jul 21, 2026, Anthropic donated another $20 million to Public First Action, bringing its total support to $40 million. On Jul 27, 2026, CEO Dario Amodei published Anthropic’s position on open-weights models. On Aug 11, 2026, Anthropic said it was supporting Governor Abbott’s Calls for Responsible Data Center Development.

Evaluation Access and Usage Rules in Late 2026

On Oct 8, 2026, Anthropic published a new version of its Usage Policy and summarized the changes in a 2026 Usage Policy update. On Oct 15, 2026, it announced an update on independent evaluators: Anthropic is giving employee-like access to a team from METR, a nonprofit that evaluates frontier AI models, and it said it was in early discussions with several other organizations, including Hugging Face. The feed also lists press contact information at [email protected] and points to support and media assets, including a downloadable press kit.

The current stage is a mixed public record: Anthropic has placed safety commitments, government-facing statements, donations, a data-center electricity commitment, regional sales restrictions, the US government directive affecting access to Fable 5 and Mythos 5, and third-party evaluation access in one chronological policy feed. The next signal is whether the METR employee-like access and early discussions with other organizations become a standing evaluation arrangement, and how the Responsible Scaling Policy Version 3.0, California compliance framework, and data-center commitments interact with the US government directive and consumer terms.

Meta description: Anthropic's 51 policy updates from 2025 to 2026 map AI safety, government access, data-center costs, export-control suspension, and METR evaluation.

Tags: AI safety, Anthropic policy, METR, data centers, export controls

Featured image: an abstract network of policy documents and neural nodes in cool blue, no real people, no company logos, no text.