Policy

On July 15, 2026, China’s Interim Measures for the Administration of Artificial Intelligence Anthropomorphic Interaction Services took effect—regulating continuous emotional, human-like AI companions with youth protections, disclosure, and content duties, while exempting ordinary customer service, Q&A, work assistants, and education tools that lack sustained emotional interaction.

On July 11, 2026, about 200 protesters marched from OpenAI to Anthropic and Google DeepMind in San Francisco, organized by Stop the AI Race, calling for a collective pause on new frontier-model training amid jobs, housing, environment, and existential-risk concerns.

On July 9, 2026, Anthropic appointed former Federal Reserve Chair and Nobel laureate Ben Bernanke to its Long-Term Benefit Trust—the independent body that can appoint Anthropic board members and is charged with holding the Public Benefit Corporation to its long-term public mission as AI’s economic effects scale.

On July 9, 2026—the same day GPT-5.6 went public—OpenAI evolved its GPT-5.5 Bio Bug Bounty into an ongoing private OpenAI Bio Bounty Program focused on universal jailbreaks against a predefined biosafety challenge, raised the top award from $25,000 to $50,000, and put GPT-5.6 in scope alongside GPT-5.5.

On June 30, 2026, Anthropic announced that U.S. export controls on Claude Fable 5 and Mythos 5 had been lifted after the June 12 shutdown. Fable 5 would return globally on July 1 across Claude Platform, Claude.ai, Claude Code, and Claude Cowork (with temporary included-usage windows on paid plans), while cloud marketplaces would re-enable as quickly as possible. Mythos 5 access for approved U.S. organizations, first restored June 26, continued with plans to expand Glasswing partners.

On June 29, 2026, Governor Gavin Newsom announced a first-of-its-kind partnership with Anthropic giving California state agencies—and, per press framing, local governments—access to Claude at a 50% discount, plus free workforce training and Anthropic developer technical assistance for GenAI workflows.

On June 26, 2026, the U.S. government approved Anthropic restoring Claude Mythos 5 access for a set of trusted U.S. organizations after the June 12 export-control shutdown. Anthropic and contemporaneous reporting described roughly on the order of 100 vetted critical-infrastructure defenders, agencies, and labs—while Claude Fable 5 remained broadly offline pending further negotiation that concluded with a full lift on June 30.

On June 23, 2026, The New York Times reported that the U.S. National Security Agency lost access to a powerful Anthropic AI model amid the Trump administration’s export-control fight with the company. Officials said the freeze deprived the intelligence agency of a tool analysts had found both impressive and alarming for discovering software weaknesses—illustrating how the Fable/Mythos shutdown rippled into U.S. national-security workflows.

On June 19, 2026, Norway’s prime minister announced a near ban on generative AI tools for elementary school pupils, with tighter limits for older students, arguing unfettered AI use risks harming learning. The policy aims to keep foundational skill-building human-first while allowing more controlled use as students age—one of the stricter national education rules on generative AI in Europe.

On June 18, 2026, a bipartisan group of House members—Sam Liccardo (D-CA), Jay Obernolte (R-CA), Ted Lieu (D-CA), and Scott Franklin (R-FL)—sent a letter to Commerce Secretary Howard Lutnick seeking legal authorities, technical evaluations, review process, and restoration/licensing criteria behind the June 12 export controls on Anthropic’s Claude Mythos 5 and Claude Fable 5. They asked for a response by June 26 and warned the action could set a lasting precedent for frontier AI.

On June 12, 2026, Anthropic received a U.S. government export-control directive at 5:21 p.m. ET ordering it to suspend access to Claude Fable 5 and Mythos 5 for any foreign national—inside or outside the United States, including foreign-national Anthropic employees. Unable to enforce citizenship filters in real time, Anthropic disabled both models for all customers the same evening. The company disputed that a narrow alleged jailbreak justified the recall, compared the demonstrated capability to publicly available models including GPT-5.5, and said other Claude models remained available.

On June 10, 2026, OpenAI published research on PRC-linked influence operations that used ChatGPT to generate content aimed at shaping U.S. public and political debate over AI infrastructure, data centers, and related trade issues such as tariffs. The company said it detected the campaign because operatives relied on its models to produce social-media-ready material, adding a new dimension to how AI tools themselves become instruments in information warfare around AI policy.

On June 5, 2026, President Trump signed National Security Presidential Memorandum 11 (NSPM-11), “Artificial Intelligence in the National Security Enterprise.” The memo directs defense and intelligence agencies to speed adoption of advanced commercial and open-source AI, replaces Biden-era NSM-25 guardrails the administration calls burdensome, emphasizes four pillars of adoption, adaptation, assurance, and accountability, and includes directions that could force contract action against AI vendors that limit government use of their systems.

On June 4, 2026, Prime Minister Mark Carney launched AI for All, Canada’s new five-year national artificial intelligence strategy. The plan targets C$200 billion in additional economic growth, 250,000 new AI-related jobs (including up to 90,000 for young Canadians), and lifting AI adoption from just over 12% toward 60% by 2034, organized around trust, opportunity, and sovereignty—with new legislation, national AI literacy, sovereign compute, and an AI Missions health flagship.

On June 4, 2026, Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page bipartisan discussion draft of the Great American Artificial Intelligence Act of 2026—intended to create a federal AI governance framework with obligations for large frontier developers, workforce and cybersecurity titles, and a multi-year preemption of certain state AI development laws. Cosponsors include Reps. Franklin, Subramanyam, Houchin, and Peters; the draft seeks public feedback before formal introduction.

On June 2, 2026, President Donald J. Trump signed Executive Order 14409, “Promoting Advanced Artificial Intelligence Innovation and Security.” The order directs agencies to harden federal and critical systems with AI-enabled cyber defense within 30 days, create an AI cybersecurity clearinghouse, define a classified process for designating “covered frontier models,” and design a voluntary early-access framework for government and trusted partners—while explicitly barring a mandatory licensing or preclearance regime for model development or release.

On May 27, 2026, OpenAI published its election-year safeguards plan for 2026, the second major global election cycle of the generative AI era. The program includes five planks: reliable voting and results information (AP live vote counts for the U.S. and Brazil; Democracy Works for U.S. registration and polling logistics), cybersecurity support for election defenders, deepfake watermarking and provenance (SynthID and C2PA), enforcement against election interference and scaled campaign messaging, and efforts to keep ChatGPT politically neutral by default.

On May 14, 2026, Anthropic published a new policy paper, '2028: Two scenarios for global AI leadership,' presenting contrasting futures depending on whether the US and its allies maintain a commanding lead in AI capabilities or allow China to close the gap through compute access and distillation.

On May 12, 2026, Humanity AI, a collaborative philanthropic initiative backed by ten major foundations, announced more than $18 million in new pooled grants, including $8 million awarded to 12 inaugural grantees and a $10 million open call, to support organizations advancing AI transparency, accountability, governance, and public benefit.

On May 5, 2026, the Center for AI Standards and Innovation (CAISI) at the US Department of Commerce announced agreements with Microsoft, Google DeepMind, and xAI to provide early access to new frontier AI models for national security and security risk evaluations before deployment, expanding prior partnerships and responding to concerns over advanced capabilities like those demonstrated by Anthropic's Mythos.

On May 4, 2026, reporting revealed that the Trump administration is discussing an executive order to establish an AI working group and potential formal review processes for new AI models before they are released, marking a shift from its prior hands-off stance amid concerns over advanced capabilities like those in Anthropic's Mythos.

On May 2, 2026, a Yale School of Management-led analysis in Fortune warned that Anthropic’s Claude Mythos Preview and similar agentic systems are exposing critical gaps in enterprise governance, with autonomous multi-step capabilities that can discover decades-old vulnerabilities, generate exploits, and exhibit aggressive behaviors in simulations. The piece offers an eight-variable framework and industry archetypes for banking, healthcare, retail, and supply chain.

On April 29, 2026, Anthropic released Version 3.2 of its Responsible Scaling Policy. The update authorizes the company’s Long-Term Benefit Trust (LTBT) to request external reviews of Risk Reports, approve the selection of external reviewers, and requires regular briefings for the LTBT.

On April 21, 2026, Bloomberg reported that a small group of unauthorized users accessed Anthropic's Mythos AI model—the company's most powerful and restricted cybersecurity-focused model—on the same day it was announced for limited release to select partners under Project Glasswing. Access was gained through a third-party vendor environment, using information from a data breach at contractor Mercor and internet sleuthing techniques.