Helen Toner Expresses Concern That ‘Inefficient’ Congress Will Mismanage AI Policy Decisions

Helen Toner, a former board member of OpenAI and the director of strategy at Georgetown's Center for Security and Emerging Technology, expressed concerns about potential "knee-jerk" reactions from Congress regarding AI policymaking if current conditions persist.

“Congress right now — I don’t know if anyone’s noticed — is not very functional and struggles to pass laws unless there’s a massive crisis,” Toner stated,“AI is a significant and powerful technology, and something will inevitably go wrong. If our only legislative responses come in reaction to drastic crises, will that truly be effective?”

Toner's remarks come just before a White House-sponsored summit on the role of AI in enhancing American innovation, shedding light on the persistent gridlock in U.S. AI policy. In 2023, President Joe Biden signed an executive order aimed at enacting consumer protections related to AI, which mandated that developers disclose safety test outcomes to pertinent government agencies. Earlier in the year, the National Institute of Standards and Technology released a roadmap for recognizing and mitigating the risks associated with AI.

However, Congress has not yet passed comprehensive AI legislation, nor has it proposed anything as extensive as the recently enacted EU AI Act. With the significant 2024 election year approaching, the likelihood of legislative progress remains low.

A report from the Brookings Institution indicates that the absence of federal regulations has prompted state and local governments to take action. In 2023, state legislators introduced more than 440% additional AI-related bills compared to 2022, with nearly 400 new AI laws proposed recently, according to the lobbying group TechNet.

California lawmakers advanced around 30 new AI-focused bills last month aimed at bolstering consumer and workforce protections. Similarly, Colorado passed a law requiring AI developers to exercise "reasonable care" to prevent discrimination in their technologies. In March, Tennessee Governor Bill Lee signed the ELVIS Act, which prohibits the AI cloning of musicians' voices or likenesses without their explicit consent.

This fragmented approach to AI regulation poses challenges and uncertainty for both industries and consumers. For example, the term “automated decision making” — which broadly refers to AI algorithms making decisions like loan approvals — is defined differently across various state laws. Some laws do not categorize decisions as "automated" if they involve any human participation, while others adopt a stricter interpretation.

Toner argues that even a basic federal mandate would be more beneficial than the current disarray.

“Some of the more thoughtful stakeholders in this space are attempting to identify sensible and straightforward guardrails we can implement now. This proactive approach could mitigate the severity of future crises and reduce the chances of needing a rapid, poorly considered response later,” she emphasized.