Washington, D.C. — Colorado U.S. Senator Michael Bennet urged U.S. Senate Majority Leader Chuck Schumer (D-N.Y.) to adopt a principles-based approach as the Senate considers writing legislation on artificial intelligence (AI). This letter comes as Schumer’s AI Insight Forums kick off today.
“Creating an AI ecosystem that fosters innovation, increases accountability, and aligns with our fundamental values should be at the top of our country’s legislative agenda,” wrote Bennet in the letter. “When considering the appropriate framework for AI regulation, we should keep in mind our continuing failure to regulate social media platforms effectively. When these platforms first emerged, they seemed to pose few material, psychological, or democratic risks…Repeating our oversight failure when it comes to more powerful technologies like AI would be disastrous.”
Bennet calls on Schumer to ensure any legislative framework reflects fundamental American values, including privacy, civil rights, and civil liberties. He highlights the need for regular public risk assessments and audits to examine the safety, reliability, security, explainability, and efficacy of AI systems, and underscores the importance of AI labeling requirements, disclosure standards, and data transparency.
“A robust AI regulatory framework will require AI developers to construct their systems so that they preserve Americans’ privacy, civil rights, and civil liberties; protect against bias and discrimination; ensure safe environments for our children; and secure the integrity of our civic processes—not as afterthoughts, but throughout the design, testing, and deployment process,” continued Bennet. “As you have made clear, this task is too significant to be left to one Member, one Committee, or Congress alone. As we design an appropriate AI framework, we will need a collaborative approach that engages AI experts and developers as well as the insights of the American people.”
Bennet has pushed strongly for greater regulation of digital platforms. In May, Bennet reintroduced the Digital Platform Commission Act, first-of-its-kind legislation to create a dedicated federal agency to oversee large technology companies. In June, Bennet called on major technology companies to identify and label AI-generated content, and introduced the Global Technology Leadership Act to bolster the government’s ability to assess U.S. capacity in emerging technologies relative to other countries. Bennet earlier introduced the Oversee Emerging Technology Act and the ASSESS AI Act to ensure government use of AI complies with fundamental rights, and joined his colleagues to introduce the REAL Political Ads Act to require a disclaimer on political ads for federal campaigns that use content generated by AI.
The text of the letter is available HERE and below.
Dear Leader Schumer:
I applaud your recent release of the SAFE Innovation Framework for Artificial Intelligence (AI). Creating an AI ecosystem that fosters innovation, increases accountability, and aligns with our fundamental values should be at the top of our country’s legislative agenda.
When considering the appropriate framework for AI regulation, we should keep in mind our continuing failure to regulate social media platforms effectively. When these platforms first emerged, they seemed to pose few material, psychological, or democratic risks. As a result, they evolved with few limitations. Now, we are intimately aware that even seemingly innocuous digital products can have deeply damaging effects on mental health, civic discourse, democratic legitimacy, and Americans’ economic agency. Repeating our oversight failure when it comes to more powerful technologies like AI would be disastrous.
As you work to develop this regulatory framework, I think it would be useful to hold several, critical elements in mind:
A Values-Based Framework. From the outset, AI regulation must incorporate our fundamental values. A robust AI regulatory framework will require AI developers to construct their systems so that they preserve Americans’ privacy, civil rights, and civil liberties; protect against bias and discrimination; ensure safe environments for our children; and secure the integrity of our civic processes—not as afterthoughts, but throughout the design, testing, and deployment process. A regulatory architecture premised on these principles will preserve cutting-edge innovation without abandoning the foundational ideals that underpin our democratic society.
Public Risk Assessments, Mitigation, and Audits. AI systems should undergo regular public risk assessments to examine their safety, reliability, security, explainability, and efficacy. We should couple these assessments with transparency and disclosure obligations to enable effective compliance audits. AI system risk assessment requirements should apply throughout the AI lifecycle, from the development stage through the post-market period, and should be conducted in an ongoing manner by AI system end-users.
AI systems that pose higher risk should undergo regular third-party audits. Organizational Audits would focus on the practices of complex AI system developers and assess a company’s governance models, corporate structure, development processes, information security, safety culture, and risk management practices. Technical Audits would be designed to detect dangerous capabilities and the propensity of higher risk AI models to cause harm.
Content Indicators. AI-generated content should retain a distinct, easily recognizable signifier, such as a watermark, hard-coded indicator, or visual overlay, so users can readily identify AI content as AI content. Platforms hosting content without such a signifier should affix their own (or remove the content).
AI Disclosure. AI platforms should disclose their AI nature at the beginning of a user’s interaction and periodically throughout in order to ensure that users understand what sort of system they are encountering.
Data Transparency. Users must understand how AI systems intend to use, store, and transfer their personal data. Users should have a right to know how their data will contribute to any AI system’s training or optimization. Users should be informed about how their data, generated by interactions with AI systems, are used. Users should have the right through “optin” procedures to determine whether AI systems can collect and use their data.
AI’s rapid development is a testament to American innovation. We should welcome its potential benefits to our economy and society. AI’s responsible deployment—consistent with American values—will require clear policies and frameworks to promote safety, anticipate risk, and mitigate harm. As you have made clear, this task is too significant to be left to one Member, one Committee, or Congress alone. As we design an appropriate AI framework, we will need a collaborative approach that engages AI experts and developers as well as the insights of the American people.
I am grateful for your leadership, and look forward to working with you on this essential project. Our future depends on getting it right.
Sincerely,