Policy

OpenAI released its Frontier Governance Framework on May 28, a formal public document explaining how its internal safety and preparedness practices map onto specific legal obligations under California's Transparency in Frontier AI Act and the EU AI Act's Code of Practice. The framework covers risk assessment for cyber offense, CBRN threats, harmful manipulation, and loss of AI control, and commits OpenAI to ongoing model reporting, incident response, and external expert review.

Beijing has expanded travel curbs to cover senior AI professionals at major private firms including DeepSeek and Alibaba, requiring pre-approval from government authorities before any overseas travel. The policy treats elite AI researchers as national security assets and escalates China's effort to prevent talent and knowledge from flowing to Western competitors.

The CMS ACCESS model, launching July 5, will for the first time allow Medicare to pay for AI-powered chronic care agents that monitor patients between visits, coordinate care, and improve measurable health outcomes. With 150+ digital health companies enrolled and two-thirds of Medicare beneficiaries in scope, it may be the most consequential federal AI policy move of 2026.

The United States has introduced more than 1,200 AI-related bills across federal and state legislatures, yet lacks a coherent national framework to evaluate whether any of them actually work. With Senator Marsha Blackburn's 291-page TRUMP AMERICA AI Act gaining momentum, states accelerating their own laws, and the White House pushing preemption that Congress already rejected 99-1, the US faces a regulatory collision course that could reshape how AI is built and deployed in America.

President Trump abruptly postponed signing a sweeping AI cybersecurity executive order hours before the ceremony, saying he 'didn't like certain aspects' of it. The order would have created a voluntary framework requiring AI labs to share frontier models with the NSA 90 days before release—and its collapse has left Washington scrambling over how to handle increasingly dangerous AI capabilities.

Colorado's legislature passed SB 189, a sweeping rewrite of the state's original AI Act that eliminates mandatory risk management programs, annual impact assessments, and algorithmic discrimination duties in favor of a narrower transparency framework. The original law was stayed by a federal court after the DOJ joined xAI in challenging its constitutionality — setting a precedent with national implications for state-level AI regulation.

Pope Leo XIV will publish 'Magnifica Humanitas,' the Catholic Church's first major teaching document on artificial intelligence, on May 25. The encyclical — signed on the 135th anniversary of Rerum Novarum — addresses the protection of human dignity in the AI age. Presenting alongside the Pope will be Christopher Olah, co-founder of Anthropic, in a pairing that signals how seriously both technology and religion are taking the moral dimensions of AI.

The Commerce Department's Center for AI Standards and Innovation has finalized pre-deployment evaluation agreements with all five major American frontier AI laboratories — OpenAI, Anthropic, Google DeepMind, Microsoft, and xAI — completing a voluntary framework under which every significant new AI model must pass government security evaluation before public release. With more than 40 assessments completed since 2024, the program is quietly becoming the de facto regulatory floor for U.S. frontier AI.

A California advisory jury unanimously rejected Elon Musk's sweeping lawsuit against OpenAI and Sam Altman on statute-of-limitations grounds, ending the highest-profile AI legal battle in history. Musk has vowed to appeal, but the ruling delivers a major legal and reputational vindication for OpenAI ahead of its rumored IPO.

The European Council and Parliament reached a provisional deal on May 7 to push the EU AI Act's high-risk system obligations back by up to 16 months, giving businesses deploying AI in hiring, education, and biometrics significantly more runway. The agreement simultaneously tightens prohibitions by explicitly banning AI used to generate non-consensual intimate imagery and CSAM.

President Trump returned from Beijing saying he and Xi Jinping discussed 'possibly working together for guardrails' on artificial intelligence, marking the first explicit US-China AI safety dialogue at the presidential level. The talks were accelerated by Anthropic's Mythos revelations and geopolitical pressure, but experts warn that neither country has a concrete framework, timeline, or enforcement mechanism.

The Nitrogen ransomware gang has claimed responsibility for a cyberattack on Foxconn's North American factories, stealing 8TB of data comprising more than 11 million files including confidential project documentation and technical drawings linked to Apple, Nvidia, Google, Intel, and Dell. The breach, which first surfaced May 1 when workers at a Wisconsin plant reported a full network collapse, is the most significant supply-chain cybersecurity incident of 2026 so far.

President Trump arrived in Beijing on May 13 for a summit with Xi Jinping, bringing a 16-CEO tech delegation that includes Apple's Tim Cook, Nvidia's Jensen Huang — a last-minute addition who joined Air Force One in Alaska — and Elon Musk. The visit puts AI chip exports, rare earths, and Taiwan at the center of the world's most consequential bilateral relationship.

Colorado's landmark AI Act — set to become the most comprehensive US state AI law on June 30, 2026 — has been stayed by a federal judge and is now being replaced by a narrower bill already passed by the state legislature. The episode reveals how fast the political ground has shifted under state-level AI regulation, and what it means for companies that spent months preparing for compliance.

As Anthropic's Mythos AI model demonstrated an unprecedented ability to discover and exploit cybersecurity vulnerabilities, America's intelligence community has launched an internal campaign to wrest AI regulatory authority away from the Commerce Department. The battle over who governs frontier AI reveals a White House increasingly uncertain about the right approach as the technology's national security implications become impossible to ignore.

The U.S. Department of War has granted AI deployment rights on its most sensitive classified networks to seven companies — Microsoft, Amazon, Google, OpenAI, SpaceX, Nvidia, and Reflection — explicitly excluding Anthropic, which was designated a national security supply chain risk in February after refusing to remove restrictions on autonomous weapons and mass surveillance use cases.

The Trump administration has invited approximately a dozen corporate CEOs — including Nvidia's Jensen Huang, Apple's Tim Cook, and Qualcomm's Cristiano Amon — to accompany President Trump on his May 14-15 Beijing visit. The invitation signals a push to reopen China's AI chip market, where Nvidia's share has collapsed to near zero following U.S. export controls and Beijing's pivot to domestic alternatives.

Apple has agreed to pay $250 million to resolve a class action lawsuit alleging false advertising over AI-powered Siri features that were marketed during the iPhone 16 launch but never shipped on time. Eligible iPhone 15 Pro and iPhone 16 owners in the US can claim up to $95 per device.

Former OpenAI board member Shivon Zilis — who also has four children with Elon Musk — testified this week that Musk attempted to negotiate a Tesla takeover of OpenAI in 2017–2018, offering Sam Altman a Tesla board seat as an inducement. The testimony adds to a growing picture of Musk's ambitions to control the AI company he co-founded.

The United States and China are preparing to add formal AI dialogue to the agenda of a mid-May summit in Beijing between President Donald Trump and President Xi Jinping — a potential breakthrough in managing the world's most consequential technological rivalry. Discussions are expected to center on preventing AI from triggering autonomous weapons escalation, curbing nonstate actor misuse of open-source models, and establishing guardrails for AI involvement in nuclear command decisions.

In overlapping moves that signal a new phase of AI governance, Google DeepMind, Microsoft, and xAI have signed formal pre-deployment security testing agreements with the Commerce Department's CAISI, while Apple, Google, and Microsoft are separately coordinating what the media has dubbed an 'AI Constitution' — a multi-layered voluntary safety framework the companies say is preferable to waiting for Congress to impose one.

NIST's Center for AI Safety and Innovation has signed pre-deployment testing agreements with Google DeepMind, Microsoft, and xAI — expanding a program that began with OpenAI and Anthropic in 2024. Under the deals, companies hand over unreleased models with reduced safety guardrails so government evaluators can assess national security risks before the public ever sees them.

Six cybersecurity agencies from the Five Eyes alliance — including CISA, NSA, and partners from the UK, Australia, Canada, and New Zealand — published their first coordinated guidance on agentic AI risks, warning that autonomous AI agents are being deployed with dangerous levels of access before adequate security frameworks exist. The document identifies expanded attack surfaces, unpredictable behavior, and threat intelligence gaps as the three primary risks of the current agentic deployment wave.

The European Union's most consequential AI compliance deadline — August 2, 2026 — is less than 90 days away, and negotiations over a proposed delay have hit a wall. A second trilogue session on April 28 ended without agreement, and most enterprises surveyed lack even a basic inventory of the AI systems they'd need to certify.

OpenAI President Greg Brockman testified Monday as the Musk v. Altman trial entered its second week, publicly confirming OpenAI's IPO plans for the first time and disclosing a personal stake worth nearly $30 billion. A bombshell filing also revealed Musk privately sought a settlement just days before filing into court.

The Trump administration is weighing an executive order that would establish a government working group to vet AI models before public release, a historic reversal after rescinding Biden's AI safety rules in January 2025. The shift was triggered by Anthropic's Mythos model — a cybersecurity AI already deployed by the NSA that the company refused to release publicly due to its offensive capabilities.

The Hangzhou Intermediate People's Court upheld a ruling that companies cannot dismiss employees solely to replace them with AI as a cost-cutting measure. The decision, published April 28 alongside similar rulings from other Chinese courts, sets a significant labor-law precedent with no equivalent yet in the US or EU, and arrives as global tech layoffs tied to AI automation topped 78,000 in Q1 2026.

The U.S. Department of Defense has approved seven major technology companies — later joined by an eighth — to deploy AI directly on its most sensitive classified networks, while conspicuously omitting Anthropic amid an ongoing legal dispute over AI safety guardrails in warfare.

The federal trial pitting Elon Musk against Sam Altman and OpenAI concluded its first week with Musk on the stand for two days, describing a nonprofit founding vision he says Altman betrayed, and clashing in heated cross-examination over a 2018 term sheet central to the $130 billion damages claim. Week two begins Monday with Altman expected to testify and a mid-May ruling anticipated from Judge Gonzalez Rogers.

Google confirmed on April 28 that it has signed a classified agreement with the US Department of Defense allowing its Gemini AI systems to be used for sensitive military operations under terms permitting 'any lawful government purpose.' The deal was announced one day after more than 580 Google employees — including senior leaders from DeepMind — signed an open letter urging CEO Sundar Pichai to refuse. Anthropic reportedly declined a similar arrangement.

Google has granted the U.S. Department of Defense access to its Gemini AI for classified networks under an 'any lawful purpose' clause — the same terms Anthropic refused, leading to the DoD designating the startup a 'supply chain risk.' Over 600 Google employees have signed an open letter urging CEO Sundar Pichai to reverse course.

The landmark civil trial between Elon Musk and OpenAI CEO Sam Altman opened April 28 in Oakland with Musk testifying he co-founded OpenAI specifically to create a nonprofit counterweight to Google. Musk is seeking $134 billion in alleged wrongful gains and Altman's removal; OpenAI argues Musk simply lost a power struggle and founded rival xAI out of spite.

China's National Development and Reform Commission has ordered Meta to unwind its $2 billion acquisition of Manus, the viral AI agent startup with Chinese roots. The NDRC cited export control and technology laws in prohibiting the deal, banned the co-founders from leaving the country, and left Meta with the messy task of disentangling a company it had already integrated into its systems.

The White House's National Policy Framework for AI, released in March 2026, recommends that Congress preempt state AI laws deemed to impose 'undue burdens.' Democrats have responded with the GUARDRAILS Act, which would block federal override. With 25 AI laws already passed in 2026 and 19 states enacting new rules in a single two-week period, the battle over who governs American AI has never been more intense.

Japan's Liberal Democratic Party has drafted legislation seeking criminal penalties for repeat offenders who use generative AI to create unauthorized images of anime and manga characters, or to produce non-consensual explicit deepfakes of real people — marking a significant hardening of the country's otherwise innovation-first AI policy and the first move toward criminal liability in the Indo-Pacific region for specific AI content misuse.

China's National Development and Reform Commission has instructed leading AI startups including Moonshot AI and StepFun to reject American investment without explicit government sign-off, Bloomberg reported Thursday. The move is a direct response to Meta's $2 billion acquisition of Manus — a deal that exposed how Chinese AI startups were restructuring offshore to bypass Beijing's oversight — and marks a significant escalation in the AI dimensions of the US-China technology war.

OpenAI held a closed-door demonstration for around 50 US cyber defense practitioners in Washington on April 22, and has begun briefing Five Eyes intelligence allies on GPT-5.4-Cyber — a permissive variant of GPT-5.4 designed for cyber defenders. The model is being offered through a new Trusted Access tiered program that separates what national security professionals can do with AI from what general consumers can.

India has established the AI Governance and Economic Group (AIGEG), a high-level inter-ministerial body chaired by IT Minister Ashwini Vaishnaw, to coordinate national AI policy across all sectors. As the third-largest AI talent pool in the world races to keep pace with the US and China, New Delhi is building the institutional infrastructure to govern — and benefit from — an AI-transformed economy.

Nvidia's $20 billion December deal with AI inference startup Groq — structured as a technology license plus mass hiring of Groq's top engineers — has drawn a Senate investigation and FTC scrutiny, with Senators Warren, Blumenthal, and Wyden arguing the arrangement is a covert acquisition designed to neutralize a competitor and entrench Nvidia's 90% GPU market dominance. Nvidia maintains the deal is a legitimate licensing agreement.

With electricity bills rising and community opposition spreading across 24 states, President Trump and major tech CEOs have signed a commitment to make AI companies pay for their own power — a political pivot forced by voter anger that has already delivered midterm-election losses to incumbents in New Jersey, Virginia, and Georgia.

Jury selection in Elon Musk's lawsuit against OpenAI begins April 27 in Oakland, California. Musk is seeking up to $134 billion in alleged wrongful gains and demanding the removal of CEO Sam Altman, while OpenAI accuses him of a last-minute 'legal ambush' that dramatically expanded the case's scope just weeks before trial.

With federal AI legislation stalled, U.S. states have become the de facto regulators of consumer AI. Nebraska's Conversational AI Safety Act passed this week, mandating chatbot disclosures and crisis protocols for minor users. Across the country, a patchwork of state-level bills is creating a fragmented compliance landscape that industry groups warn could harm innovation — while advocates argue it is the only protection consumers have.

A bipartisan group of U.S. lawmakers has introduced the MATCH Act, legislation that would ban exports of deep-ultraviolet (DUV) immersion lithography systems and related chipmaking equipment to Huawei, SMIC, CXMT, YMTC, and other Chinese firms. The bill also pressures allied nations — including the Netherlands, Japan, and South Korea — to align their own export controls within 150 days or face U.S. sanctions.

OpenAI, Anthropic, and Google have activated the Frontier Model Forum as a live threat-intelligence operation, pooling defenses against an industrial-scale adversarial distillation campaign traced to DeepSeek, Moonshot AI, and MiniMax. Anthropic alone documented 16 million unauthorized extractions via 24,000 fake accounts and has banned all Chinese-controlled companies from Claude.

Maine's LD 2082, which bans licensed mental health providers from using AI for independent therapeutic decisions, passed both legislative chambers on April 7–8 and heads to the governor. Missouri's HB 2372 passed the full House on April 2 with a $10,000 first-offense penalty clause, and now sits in the Senate. The twin developments signal an accelerating state-level movement to draw hard legal lines around AI in mental healthcare.

OpenAI has published a sweeping 13-page policy document urging governments to prepare for approaching superintelligence by taxing automated labor, creating public wealth funds seeded by AI companies, and piloting a 32-hour workweek with no pay cuts. Released days before a likely IPO roadshow, the proposals position OpenAI as a responsible steward of economic disruption — while critics note the company stands to benefit from the regulatory frameworks it is proposing.

Utah's regulatory sandbox has approved Legion Health to allow its AI chatbot to autonomously renew prescriptions for 15 psychiatric medications—making Utah the first government in the world to authorize AI for autonomous psychiatric prescribing. The approval comes weeks after a previous Utah AI prescription pilot was successfully jailbroken, raising immediate questions about the clinical safety of moving faster than medical safeguards can keep up.

Governor Newsom signed Executive Order N-5-26, requiring AI vendors to disclose safety policies on CSAM, civil rights, and anti-discrimination as conditions for state contracts. The order also grants California the authority to independently overrule federal AI supply-chain risk designations, setting up a direct constitutional clash with the Trump administration.

Governor Gavin Newsom signed Executive Order N-5-26, establishing first-in-the-nation AI procurement standards for California state contracts and explicitly decoupling the state's supply chain risk assessments from federal designations. The move escalates the battle between California and the Trump administration over who gets to govern American AI.

Three of the world's leading AI labs are formally sharing intelligence through the Frontier Model Forum to detect and block 'adversarial distillation' — a technique where outside labs bombard US AI systems with automated prompts to clone their behavior. Anthropic alone documented 16 million extraction attempts linked to Chinese AI firms, escalating the fight over intellectual property at the AI frontier.

The DOJ's AI Litigation Task Force, operational since January 10, is now actively challenging state AI statutes that conflict with the Trump administration's December 2025 executive order preempting local regulation. With over 20 states having enacted comprehensive AI laws, the outcome of this federal-state standoff will define who governs AI in America for the next decade.

Governor Kathy Hochul signed New York's Responsible AI Safety and Education Act into law on March 27, establishing the most stringent AI safety framework enacted by any U.S. state. The law takes effect January 1, 2027, but the Trump administration's DOJ AI Litigation Task Force is already positioned to challenge it — setting up a defining constitutional battle over who governs AI in America.

Bipartisan legislation introduced April 2 would require US allies — including the Netherlands and Japan — to restrict deep ultraviolet lithography equipment exports to China, and threatens extraterritorial US jurisdiction if they don't comply. The bill specifically targets ASML and Tokyo Electron equipment still flowing to Huawei and SMIC.

Brussels isn't bluffing. Three companies received formal notices this week under the AI Act's transparency provisions. The real question is whether this helps or hurts European AI competitiveness.

Three landmark cases are working through US courts right now. The outcomes will determine whether AI companies owe billions in licensing fees — or whether training on public data remains fair use.

New export restrictions just dropped, and NVIDIA is caught in the middle. They're losing $15B+ in China revenue while Beijing accelerates domestic chip development. Nobody wins this game.