Home · Politics · Oct 7 archive
Cantwell Proposes Federal Testing Role in AI Safety Framework
Confirmed
In Short: Cantwell's framework, which she presented on Oct. 15, calls for public-private partnerships to develop defensive AI tools that can detect malicious activity, patch vulnerabilities, and protect critical infrastructure.
Sen. Maria Cantwell, D-Wash., is proposing a federal role in testing and regulating advanced AI systems to ensure safety and security. Cantwell outlined a comprehensive AI governance framework that includes continuous testing by government and independent experts, along with independent audits to verify that developers meet federal safeguards before releasing covered models.
Cantwell's framework, which she presented on Oct. 15, calls for public-private partnerships to develop defensive AI tools that can detect malicious activity, patch vulnerabilities, and protect critical infrastructure. She also advocates for human oversight in areas such as employment, healthcare, and credit, and a process for people to challenge AI-influenced decisions.
The senator emphasized the need for clear safety standards, continuous testing, and reporting of serious failures in AI systems. She stated, “America can lead the world in AI by building systems that are not only more capable, but safer and more secure.”
Cantwell's framework stands in contrast to the White House and some congressional Republicans' focus on voluntary standards and self-regulation. It calls for mandatory independent safety testing and reporting of incidents involving dangerous cyber capabilities, failures of critical safeguards, or loss of control over autonomous agents.
The framework also addresses the need for international cooperation on common AI safety standards and information-sharing mechanisms. Cantwell suggests a secure channel with China similar to the Soviet-era red phone for rapid communication during significant AI incidents.
According to a Democratic committee aide, the framework could develop into more than one bill, with the intention of developing bipartisan legislation. The aide noted that enforcement mechanisms, such as civil or criminal penalties, will be determined during discussions.
Cantwell's proposal comes as other tech leaders and policymakers are also calling for more stringent AI regulations. On Sept. 29, former President Trump announced a voluntary safety agreement with several tech companies, while OpenAI called for mandatory, ‘capability-based’ national regulation.
Two years ago, Cantwell advocated for five bipartisan AI measures included in a larger package of artificial intelligence legislation. She reintroduced one of these measures, the Future of AI Innovation Act, in February.
The framework aims to advance the public discussion on AI governance and ensure that AI systems are not only more capable but also safer and more secure.
What's confirmed
- “To manage risks from advanced AI systems we need clear safety standards, continuous testing, and reporting of serious failures,” Cantwell said in a statement.
- “America can lead the world in AI by building systems that are not only more capable, but safer and more secure.” Last month, Cantwell took to the Senate floor to demand urgent federal regulations and mandatory independent safety testing for frontier AI models before their release.
What's still developing
- On Sept. 29, Trump announced a voluntary safety agreement with Nvidia, SpaceX, OpenAI, Anthropic, Meta and Google calling for “robust internal controls” to prevent unauthorized hacking and to work with independent external auditors.
- Gavin Newsom signed Senate Bill 813 and Assembly Bill 1405, establishing the first state framework for independent assessment and verification of AI safety and risks.
- Maria Cantwell, D-Wash., ranking member of the Senate Committee on Commerce, Science and Transportation, on Oct.
- While Senate Majority Leader John Thune, R-S.D., has said he’s working on an AI-focused bill, Cantwell is not involved in negotiating that legislation.
- The enforcement mechanism “will have to come out during discussions as we turn this into a bill and through discussions with partners as we bring on others on board,” the committee aide said, but it could include civil or criminal penalties based on the particular violation.
- They would also be required to report safety incidents, including “unsafe recursive self-improvement” in which an AI model develops itself or other models without human instruction.
- “Standards should establish proportionate safeguards and risk-management requirements, with heightened review and oversight for systems that exhibit dangerous capabilities or are capable of self-improvement that may bypass safety controls,” the framework said.
- “If you’re testing an AI model to diminish its ability to design a biological weapon or a nuclear weapon, that might be something where you need national security expertise,” the aide said.