Congress unveils bipartisan framework to regulate AI

A bipartisan group of lawmakers on Monday introduced a proposed framework aimed at regulating advanced artificial intelligence systems, seeking to address safety, transparency and national security concerns as AI capabilities expand rapidly. The proposal outlines new requirements for testing high-risk systems, mandatory disclosures about training data and model capabilities, and measures to bolster federal oversight and industry compliance.

Proponents say the framework is designed to balance innovation and public safety, drawing on recent executive branch guidance and international regulatory developments. The proposal arrives amid growing public scrutiny of AI following high-profile incidents involving misinformation, privacy breaches and automated decision-making in hiring, lending and criminal justice contexts.

Background and history of the issue

Interest in federal AI regulation has intensified over the last several years as commercial AI systems grew more powerful and more widely deployed. The European Union adopted the AI Act in 2023, creating a tiered risk-based approach that has influenced debates elsewhere. In the United States, the White House issued an executive order on AI in 2023 that emphasized safety, security and civil rights safeguards while encouraging industry innovation.

Until now, federal oversight of AI has been largely sectoral, with agencies such as the Federal Trade Commission, the Department of Health and Human Services and the Department of Transportation applying existing consumer protection and safety statutes to specific uses. Lawmakers from both parties have increasingly called for a more coordinated, cross-cutting regulatory approach to address risks that cut across sectors.

Key facts, figures and stakeholders

The framework unveiled by the bipartisan group sets out several core elements intended to form the basis of legislation. Among the main proposals are:

• Mandatory independent safety testing for AI systems deemed high risk before deployment.
• Requirements for companies to publish clear disclosures about an AI system’s capabilities, known limitations and the types of data used for training.
• New authorities for federal agencies to audit and, when necessary, restrict or pause AI systems that pose imminent risks to public safety or national security.
• Funding for technical expertise within federal agencies to enable effective oversight and enforcement.
• Incentives for the development of open, interoperable standards and support for AI safety research.

Officials involved in drafting the proposal said it is intended to be technology neutral and scalable, applying strongest controls to systems that have the greatest potential for harm. They described the approach as emphasizing accountability rather than bans on research or deployment.

Dozens of lawmakers from both parties are reported to have expressed early support for the outline, and several industry groups said they were willing to engage in negotiations. Major technology companies have indicated they would prefer a federal framework over a patchwork of state laws, while some civil society organizations argued the proposal should include stronger enforcement mechanisms and clearer protections for privacy and civil rights.

Reactions

Analysts say the bipartisan nature of the proposal increases its chances of advancing through Congress, but they also caution that substantial disagreements remain over scope, enforcement and the balance between innovation and regulation. “Bipartisan frameworks can set the floor for more specific rules, but translating principles into law will require hard choices,” an analyst at a policy think tank noted.

Industry representatives voiced cautious optimism. Several trade groups welcomed federal leadership and emphasized the importance of regulatory clarity for continued investment. “Clear rules will help companies invest safely and compete globally,” a trade association spokesperson said.

Advocacy groups and civil rights organizations offered mixed responses. Some praised the focus on high-risk systems and transparency, while others urged stronger protections, including independent audits, stronger privacy safeguards and robust avenues for redress for people harmed by AI decisions.

Broader implications and what happens next

If lawmakers move from framework to formal legislation, the next steps would likely include committee hearings, markups and negotiations among House and Senate authors. Key hurdles include agreeing on enforcement mechanisms, penalties, funding levels for agency oversight and how to define the systems that would be considered high risk.

Experts say the framework could have significant global implications. A U.S. law that balances safety and innovation may influence regulatory approaches in other countries and affect how multinational companies design and deploy AI systems. The interplay between a potential U.S. law and the EU AI Act will be closely watched, especially for companies operating across borders.

Economists and technology policy specialists caution that poorly designed rules could slow beneficial innovation or concentrate compliance costs among smaller firms. Supporters argue that targeted, risk-based rules will protect the public while allowing beneficial uses of AI in medicine, education and economic productivity.

Frequently Asked Questions

What does the proposed framework require of AI companies?

The framework calls for safety testing of high-risk systems, public disclosures about capabilities and training data, and compliance with agency audits. It also proposes funding to build federal technical expertise for oversight.

Who would enforce the rules under this framework?

The proposal envisions a combination of existing federal agencies taking on enforcement roles along with a potential coordinating office to oversee cross-agency work. Specific enforcement authorities and penalties would be negotiated during the legislative process.

How does this differ from the European Union’s AI Act?

The EU AI Act, adopted in 2023, established a comprehensive, risk-based regulatory system with detailed obligations and penalties. The U.S. framework emphasizes a principles-based, technology-neutral approach with sectoral agency enforcement and aims to balance innovation and safety.

Will the framework affect small startups?

The drafters say the framework is intended to be scalable, applying the strictest rules to systems that pose the highest risk. Lawmakers and stakeholders will need to address how compliance costs are distributed to avoid undue burdens on smaller firms.

How long before any law could take effect?

Timelines vary. If legislation is introduced and moves through both chambers, it could take many months to a year or more before a law is enacted and agencies have time to develop implementing rules. The process will depend on legislative priorities and negotiation outcomes.

What role will public input play?

Public comment periods, hearings and technical consultations are common parts of the rulemaking process. Stakeholders including civil society, industry, academia and state governments are expected to provide feedback during drafting and implementation stages.

Could this framework affect international AI development?

Yes. A major U.S. law could shape global industry standards and influence regulatory approaches in other countries, particularly if provisions align with or differ from rules in the EU and other jurisdictions.

Supporters of the framework argue that federal action is needed to provide legal clarity and protect the public as AI systems become more integrated into daily life. Opponents caution that without careful design, regulation could hamper innovation or fail to prevent harm. The coming weeks and months will determine whether the bipartisan outline evolves into formal legislation and what form any final law will take.