Passing The Buck
The hardest part of regulating AI in the West appears to be deciding who will take responsibility.
The EU began enforcing its AI Act on August 2, activating transparency obligations, general-purpose AI oversight, and a penalty regime two years in the making. Across the Atlantic, the most ambitious federal AI bill in American history failed to reach a floor vote before Congress left for August recess. The UK’s promised AI legislation never reached Parliament. Three governments, working from different starting points, arrived at remarkably similar difficulties within weeks of one another.
Political deadlock
Representatives Jay Obernolte and Lori Trahan released a 269-page bipartisan discussion draft on June 4, titled the Great American Artificial Intelligence Act of 2026. The bill addressed frontier model transparency, codified the Center for AI Standards and Innovation at NIST, and included cybersecurity and workforce provisions. Its centerpiece was a three-year preemption of state laws that specifically regulate AI model development, a provision that the sponsors framed as necessary to prevent a patchwork of fifty conflicting regulatory regimes.
The preemption clause drew simultaneous opposition from both parties. Democrats and civil society groups, including Public Citizen and the AFL-CIO, argued that a federal freeze would strip states of tools already in use to address AI harms. Republicans and industry groups pushed for broader preemption, contending that the freeze should extend to deployment-side state laws. Neither faction found the bill acceptable as drafted, and no coalition formed around a compromise. The Great American AI Act was never formally introduced. Congress entered August recess with no federal AI statute, while California and Colorado enforce state-level AI rules in the vacuum.
Nobody wants the job
The UK government committed to “appropriate legislation” for frontier AI models during 2024 and reaffirmed the promise into 2025. When the May 2026 King’s Speech announced 37 bills for the parliamentary session, none addressed artificial intelligence. Ministers had concluded that AI should be regulated at the point of use by existing sector regulators. In January 2026, the Department for Science, Innovation and Technology asked 19 regulators to publish individual plans for handling AI within their existing mandates, distributing authority across the ICO, Ofcom, the FCA, and others. Adjacent legislation contributed as well. The Data (Use and Access) Act 2025, which took effect in February, reformed automated decision-making rules, and the Crime and Policing Act 2026 addressed AI-generated deepfakes. Both imposed binding requirements on AI systems through existing legal channels, filling regulatory gaps without a dedicated AI statute.
Lord Holmes of Richmond’s Private Member’s AI Regulation Bill, which proposes a dedicated AI Authority to coordinate across regulators, passed the House of Lords and reached the Commons without government support. Parliament demonstrated appetite for a new institution to own AI governance, but the executive preferred to avoid creating one, choosing instead to layer AI responsibilities onto bodies that already exist.
Bureaucratic idealism
The transparency obligations that took effect on August 2 cover chatbot disclosure, machine-readable marking of AI-generated content, and deepfake labeling. Enforcement powers over general-purpose AI models also activated, granting the AI Office authority to issue information requests, access models, and order recalls. The full penalty regime arrived with them. Among the three jurisdictions, the EU alone reached the stage of active enforcement, two years after the AI Act entered into force.