AI Regulation News: California AI Disclosure Order, EU AI Act Delays, Oregon Chatbot Law — April 7, 2026

April 7th 2026

The states are building while the feds are still drafting blueprints.

A seismic week for AI governance on both sides of the Atlantic. California's governor signed an executive order asserting state authority over AI procurement just days after the White House called on Congress to preempt state-level AI laws. Meanwhile, the European Parliament voted overwhelmingly to delay high-risk AI obligations, and Washington State quietly became the latest to pass frontier model safety legislation.

Watch & Listen First

  • The AI Policy Podcast: Inside Project Maven and AI-Powered Warfare -- CSIS's Gregory Allen interviews Katrina Manson on the intersection of military AI deployment and the governance gaps it exposes. (March 26 episode, Spotify)

  • RegulatingAI Podcast: Innovate Responsibly -- Sanjay Puri's weekly conversations with global leaders wrestling with the innovation-vs-safety tradeoff. Fresh episodes weekly. (Spotify)

  • EU AI Act Podcast: Decoding AI Regulation -- RegInt's deep dives into the legal mechanics of Europe's shifting AI compliance landscape, directly relevant as Parliament votes to delay. (Spotify)

    • Key Takeaways

  • The federal-vs-state AI turf war is now open conflict. The White House AI framework explicitly calls for preempting state laws, while Newsom's executive order reserves California's right to overrule federal AI risk designations. This is heading to court.

  • Europe is blinking on its own rules. The EU Parliament voted 569-45 to push high-risk AI obligations from August 2026 to December 2027, and Amnesty International warns the "simplification" agenda is gutting protections, not streamlining them.

  • AI money is rewriting midterm politics. With over $185 million already deployed by AI industry players and dueling super PACs spending nine figures each, the 2026 midterms are becoming a proxy war over who regulates AI.

  • State legislatures are moving faster than anyone expected. Washington signed frontier model safety into law, 78 chatbot bills are alive across 27 states, and Georgia has three AI bills on the governor's desk. The patchwork the White House fears is already here.

  • 78% of enterprises are not ready for the rules already on the books. A new readiness report finds the vast majority of organizations have not taken meaningful steps toward AI Act compliance, even as deadlines approach.

    • The Big Story

    Newsom Signs Executive Order Asserting California's AI Authority Over Federal Government . March 30 . CalMatters → Governor Newsom signed Executive Order N-5-26, requiring AI vendors seeking state contracts to disclose their usage policies, civil rights safeguards, and content moderation practices. The order's most provocative provision gives California's Chief Information Security Officer authority to overrule federal designations of AI companies as supply chain risks -- a direct challenge to the Trump administration's push for centralized federal control. With a 120-day clock now ticking for new certification standards, California is building the regulatory infrastructure the White House explicitly wants to preempt.

    Also This Week

    White House Releases National AI Policy Framework Calling for Federal Preemption . March 20 . WilmerHale → The framework wants Congress to block states from regulating AI model development or imposing developer liability -- but Congress has already rejected preemption twice.

    EU Parliament Votes 569-45 to Delay High-Risk AI Act Obligations . March 26 . European Parliament → Annex III high-risk systems get until December 2027; Annex II systems until August 2028, because harmonized technical standards still aren't ready.

    Washington Governor Signs Frontier AI Safety Legislation . April 3 . Transparency Coalition → SB 3312 creates the AI Safety Measures Act for large frontier models, making Washington one of the first states to regulate foundation model risk directly.

    AI Super PAC Raises $125M for Midterm Spending . March 2026 . ABC News → Backed by OpenAI, a16z, and Palantir founders, "Leading the Future" is pushing candidates who favor a single federal framework over state-level regulation.

    Amnesty International Warns EU "Simplification" Will Gut AI and Privacy Protections . April 2026 . Amnesty International → The Digital Omnibus would let companies skip removing personal data from AI systems if it requires "disproportionate effort," effectively carving out GDPR exemptions for AI training.

    Deadlines & Compliance

  • June 30, 2026: Colorado AI Act takes effect. Developers must begin exercising "reasonable care" against algorithmic discrimination; deployers must complete initial impact assessments. (Clark Hill)

  • July 28, 2026 (approx.): California's 120-day window closes for new AI procurement certification standards under Executive Order N-5-26.

  • August 2, 2026: EU AI Act high-risk obligations originally take effect -- but the Parliament vote to delay to December 2027 is now in trilogue negotiation. Plan for August; hope for December. (Kennedys Law)

    • Worth Reading

  • White House AI Framework Full Text (PDF) -- The 18 pages that launched the preemption war. Read it before the pundits summarize it for you.

  • Vision Compliance: 78% of Enterprises Unprepared for AI Act -- The numbers behind the delay vote -- most companies aren't close to ready.

  • How EU Proposals Will Roll Back Rights to Feed AI -- Amnesty's detailed breakdown of what the Digital Omnibus actually changes.

  • California AI Rules Set National Testing Ground -- Axios on why Sacramento, not Washington, is setting the pace.

    • The White House wants one rulebook. The states are writing dozens. The EU can't finish the one it already passed. Somewhere in this mess, compliance teams are updating their spreadsheets and hoping for clarity that isn't coming.