Eleven AI Bills Stalled in House Committees

Eleven bills forming a comprehensive artificial intelligence framework have passed the House Science Committee, yet not a single one has been scheduled for a floor vote. Rep. Jay Obernolte, a Republican from California, indicated that the legislation could be consolidated into a single measure by December. However, the path to that vote remains blocked by procedural hurdles and ongoing debates over federal versus state authority.

The original proposal, known as the Great American AI Act, was a 269-page draft covering research funding, cybersecurity, and safety standards. To navigate congressional jurisdiction, lawmakers split it into eleven separate bills. This fragmentation has slowed progress, with critics describing the process as frustratingly slow. The stakes are high, as the package includes the first major federal attempt to regulate the most powerful AI systems.

Federal Preemption Sparks State Opposition

The most contentious element of the package is a provision that would temporarily override state laws governing AI model development. For three years, unless renewed by Congress, federal rules would take precedence over state regulations regarding how models are built. This preemption would invalidate specific California laws requiring training data transparency and content watermarking. While federal uniformity is often cited to reduce compliance burdens, it effectively freezes state-level innovation in AI governance.

Supporters argue that a single federal standard prevents a patchwork of conflicting rules that could stifle innovation. Opponents, however, warn that this approach could dilute consumer protections already enacted in states like California, New York, and Illinois. The trade-off is clear: national consistency comes at the cost of local regulatory agility and existing safety mandates.

Safety Rules Target Frontier Models

The FRONTIER Act, a key component of the package, introduces tiered safety requirements based on a system’s capabilities. Companies training the largest and most capable models would face strict obligations, including mandatory incident reporting, risk management plans, and independent audits. In contrast, firms that merely use existing AI tools would face lighter regulations. This approach aims to focus regulatory resources on systems with the highest potential for catastrophic harm.

According to reporting by GN technics/ai (en-US), both major AI developers and safety advocates have found common ground on these transparency and auditing measures. This overlap is seen as a potential fast track for passing specific provisions. However, until the full package is resolved, these critical safety standards remain pending, leaving a gap in federal oversight for high-risk AI systems.

Bipartisan Efforts Face Procedural Delays

Despite bipartisan support for key safety provisions, the legislation has not reached the House floor. Rep. Sam Liccardo, a Democrat from California, emphasized the need to finalize the package, noting that serious AI legislation has been lacking in recent years. The delay highlights a broader challenge in Congress: balancing urgent technological risks with the slow mechanics of legislative compromise.

As the House returns in November for potential hearings, the outcome of these bills will determine whether the US establishes a new federal baseline for AI safety or leaves regulation fragmented across state lines. The catch is that without a floor vote, the status quo of inconsistent state laws and undefined federal responsibilities will persist, potentially complicating the development and deployment of next-generation AI systems.