Congress's backlog of AI legislation: treat cancer, regulate chatbots.

      TL;DRUS legislators unveiled a group of AI-related bills around the July 4th holiday, representing competing perspectives. Some propose using AI for pediatric cancer research, detecting tax fraud, border monitoring, and federal prize competitions. Others aim to impose restrictions, including mandatory chatbot protections for minors, labeling of AI-generated material, and requiring federal certification for data centers. While few of these measures are likely to become law, they reveal the growing political pressure surrounding AI.

      In a span of approximately ten days, U.S. lawmakers submitted a significant amount of legislation regarding artificial intelligence. The bills, compiled by Nextgov/FCW, reflect two contrasting approaches. One group views AI as a tool for the government to utilize more rapidly, while the other regards it as a potential danger that requires regulation. This conflicting sentiment was evident within the same week and from the same legislative body, highlighting the central tension at play.

      AI as a solution

      On July 9, Representative Michael McCaul put forth the Accelerating Innovation for Kids with Cancer Act, aiming to appoint an AI Innovation Coordinator to guide federal support for pediatric cancer research. The coordinator's role would be practical, focusing on preparing cancer research data infrastructure for AI analysis and enhancing clinical trial designs to accommodate varied data types.

      Additionally, Representative Vern Buchanan proposed the AI Tax Integrity Act, which would establish a pilot program within the Treasury to use AI to detect identity theft and fraudulent tax refund claims, running for 18 months to two years. Representative David Schweikert proposed utilizing anomaly-detection algorithms at Arizona's border crossings, while Representative Nick Begich suggested integrating a federal prize challenge into the National AI Initiative Act, covering diverse areas from materials science to mechanistic interpretability.

      AI as a problem

      In contrast, the opposing bills present more straightforward restrictions. The People-First Chatbot Act, introduced by Representatives Valerie Foushee and Greg Casar, seeks to prevent companies from training chatbots using conversations from underage users. It would also mandate companies to disable harmful features for minors, clearly indicate when users are interacting with a machine, and conduct monthly safety evaluations. Users would have the right to access and delete their stored chat logs.

      The necessity of these measures is underscored by recent incidents linked to AI chatbots, with Foushee drawing parallels to legal actions against Big Tobacco. Notably, Florida is suing OpenAI and has named Sam Altman personally in relation to ChatGPT's safety, indicating a legislative response to both litigation and lobbying efforts.

      Labels, quantum, and the Markey package

      On July 2, a bipartisan group led by Representative Josh Gottheimer introduced the Spot the Fakes Act, which would mandate the FTC and NIST to establish technical regulations that require AI-generated content to include a label in its metadata. Gottheimer emphasized that the public deserves to know if content is machine-made, as the industry has been gradually moving in this direction, with OpenAI adopting C2PA metadata and SynthID watermarks.

      Gottheimer also collaborated with Representative Mike Lawler on the Advancing American Quantum Leadership Act, aiming to broaden an Export-Import Bank initiative from "quantum computing" to the wider "quantum information science and technology" category, positioned as a competitive move against China.

      Senator Ed Markey has also introduced an AI Accountability Agenda, compiling ten existing bills that address issues concerning worker rights, child privacy, algorithmic bias, healthcare, data center impacts, and equitable sharing of AI benefits.

      The data center debate shifts to the House

      Markey's initiative includes the proposed Protecting Communities from Data Center Impacts Act, which would require data centers to acquire federal certification before construction, demonstrating that they meet essential energy, environmental, and economic benchmarks. Although he has not officially introduced it in the Senate, Representative Greg Landsman has presented it in the House, advocating for community awareness of environmental impacts.

      Landsman has previously addressed this issue, as the House has already considered his bill aimed at reallocating AI data center energy costs back to the companies responsible for them. This development is becoming a consistent source of local political concern.

      What happens next

      It is unlikely that most of these bills will be enacted. Many proposed legislations typically fail to progress out of committee, and a collection of introductions often reflects anxiety rather than concrete policy proposals. However, this situation illustrates where political pressure is increasing, particularly concerning child safety, content labeling, and data center locations. The ongoing conflict remains unresolved, as Congress seeks to advance AI rapidly enough to address health challenges while also ensuring it is regulated adequately to protect public safety—a balance that has yet to be achieved.