AI Summit_Sept. 13 2024

Congressional Research Service

The Artificial Intelligence Consumer Opt-in, Notification, Standards, and Ethical Norms for Training Act , or the AI CONSENT Act (S. 3975), was introduced on March 19, 2024. The bill would require the FTC to promulgate regulations to prohibit covered entities from using, or selling or transferring to a third party, any covered data of an individual that is collected by the covered entity to train an AI system unless the entity first (1) provides a clear and conspicuous disclosure of how the entity or a third party will use the covered data, and (2) obtains express informed consent from the individual. The bill specifies that such regulations would have to require that such consent may be easily revoked at any time, obtained independently from the terms of service agreement, and cannot be inferred from closing a window or piece of content, that services cannot be conditional on consent, and that if an individual revokes consent, all of the covered data must be expunged from datasets subsequently used to train AI systems. The Federal AI Governance and Transparency Act (H.R. 7532) was introduced on March 5, 2024. The bill would amend 44 U.S.C. Chapter 35 ² Coordination of Federal Information Policy, by adding a new Subchapter IV, which would require the Director of OMB to oversee the design, development, acquisition, use, management, and oversight of federal AI systems by agencies in meeting the purposes of the bill, including ensuring the safety, security, resiliency, transparency, and accountability of federal AI systems. The bill would also require the head of each federal agency to undertake various responsibilities with respect to federal AI system management, policies, and procedures, including maintaining a publicly available webpage with information about agency policies and procedures for governing AI systems, and modifying (as necessary and appropriate) the agency appeals process to account for determinations made or substantively informed by a federal AI system. Additionally, the bill would require each agency head to oversee the establishment of an AI governance charter, in accordance with OMB guidance, and have a publicly available AI governance charter for each federal $, V\VWHP ³WKDW LV GHVLJQDWHG DV D KLJK -risk federal [AI] system or was trained on, uses, or produces a record maintained on an LQGLYLGXDO ´ The bill would require the General Services Administration to maintain a central inventory of the AI governance charters, and the Inspector General of each agency to conduct periodic independent evaluations of the AI governance policies and practices. The AI Foundation Model Transparency Act of 2023 (H.R. 6881) was introduced on December 22, 2023. The bill would require the FTC ² in consultation with NIST, OSTP, and the U.S. Copyright Office ² to establish standards specifying information to improve the transparency of foundation models provided by covered entities with respect to training data, model documentation, data collection in inference, and operations of foundation models. The FTC would also be required to issue guidance for agencies, and to consider whether to include alternate provisions in the standards for open-source foundation models, or foundation models that are derived from or built upon another foundation model. The Eliminating Bias in Algorithmic Systems Act of 2023 (S. 3478) was introduced on December 12, 2023. The bill would require covered federal agencies that use, fund, or oversee algorithms to have an office of civil rights focused on bias, discrimination, and other harms of covered algorithms, including but not limited to computational processes using AI techniques. The bill would also require that the civil rights office of each covered agency periodically report to congressional committees with jurisdiction over the agency on the state of the field and technology of covered algorithms with respect to the DJHQF\¶V jurisdiction, any steps taken to mitigate harms from the algorithms, actions to engage stakeholders, and any recommendations for legislation or administrative action to mitigate bias, discrimination, or harms from covered algorithms. Additionally, the Civil Rights Division at the Department of Justice would be required to establish an interagency working group on covered algorithms and civil rights.

AI Roundtable Page 190