TAG Act
- Sponsor
- Sen. Peters, Gary C. [D-MI]
- Committees
- Homeland Security and Governmental Affairs Committee (primary)
- Last action
- Sep 30, 2026
Bottom line
The TAG Act establishes a comprehensive framework for transparency, accountability, and privacy in federal agencies' use of AI and automated systems, particularly in decisions affecting individuals.
What it actually does
This bill directs the Director of the Office of Management and Budget (OMB) to issue guidance requiring federal agencies to be transparent about their use of artificial intelligence (AI) and automated systems when interacting with the public or making critical decisions. It mandates public inventories of AI systems, plain language notices to individuals, robust appeal processes for adverse decisions, and opportunities for human review. The bill also includes strict provisions on data privacy, prohibiting commercial use of data collected by agency AI systems and preventing agencies from requiring private account creation for public interaction.
Proponents argue
Proponents argue that this bill is crucial for maintaining public trust in government operations as AI adoption grows, ensuring that individuals are aware when AI influences decisions affecting their lives and providing clear avenues for appeal and human oversight. They contend it protects civil rights and liberties, promotes due process, and establishes responsible AI development and deployment within the federal government.
Opponents contend
Opponents might argue that the extensive requirements for public inventories, detailed notices, appeal processes, and human review could impose significant administrative burdens and costs on federal agencies, potentially slowing down the adoption of beneficial AI technologies. They may also raise concerns about the practicality of providing human review for every critical decision influenced by AI, or that over-regulation could stifle innovation.
The bill is relatively concise and well-structured, making it accessible for an informed reader to understand its core provisions within a reasonable timeframe, allowing for a thorough evaluation.
Section 3(a)
Public Inventory of Augmented Critical Decision Processes
This provision requires the Director of OMB to issue a directive within 90 days of the Act's enactment. This directive will compel each federal agency to publish a public inventory within 180 days of enactment, detailing all augmented critical decision processes, the specific use of artificial intelligence within those processes, and identifying any contracts and vendors involved. Agencies are also required to provide biannual public notifications while the comprehensive guidance is being finalized.
Supporters argue
Supporters argue this provision is foundational for accountability, allowing the public, oversight bodies, and researchers to understand where and how AI is being used by the government, which is essential for informed public discourse and building trust.
Critics contend
Critics might argue that creating and maintaining such comprehensive inventories, especially with vendor details, could be resource-intensive for agencies and potentially expose sensitive operational information, though the bill focuses on transparency for public-facing critical decisions.
Tradeoffs
This provision balances the public's right to know about government AI use with the potential administrative overhead and resource allocation required for agencies to compile and maintain detailed inventories.
Section 3(b)(5)-(7)
Plain Language Notice, Appeal, and Recourse for Adverse AI Decisions
This section mandates that agencies provide clear, conspicuous, and prompt plain language notice to individuals when a critical decision concerning them was made using an augmented critical decision process. The notice must include sufficient information to enable individuals to appeal the adverse decision and obtain timely recourse. Agencies are required to establish a public appeals process, publish it on their website, set a maximum number of days for the appeals process, and implement a system for individuals to track their appeal requests.
Section 3(b)(9)
Opportunity for Human Review and Information on AI Decision Processes
This provision requires agencies to provide individuals who are subject to critical decisions made by an augmented critical decision process with two key opportunities: first, an alternative review by a human being working for or on behalf of the agency, independent of the augmented process; and second, an opportunity to request information about the augmented critical decision process used, including the data, information, and process that led to the decision.
Section 3(c)(3)-(5)
Transparency and Human Escalation for AI Chatbots and Similar Systems
This section mandates that agencies using AI systems, such as chatbots or automated voice agents that mimic human interaction, provide prompt, plain language notice to individuals that they are interacting with an AI system, with reiterations during extended interactions. Agencies must also establish a process for individuals to timely escalate requests or issues to a human being. Furthermore, agencies must provide plain language notice on how any personally identifiable information (PII) input into the AI system will be used, ensuring compliance with data privacy laws and explicitly prohibiting transmission of PII to private companies.
Section 3(f)-(h)
Data Use Restrictions and Account Requirements for Agency AI Systems
This section establishes several prohibitions: it forbids contractors or third-party vendors from using data collected through agency AI systems for commercial purposes (e.g., targeting advertisements, creating personal profiles beyond contract performance). It also prohibits agencies from requiring the public to create private or commercial accounts as a condition of interacting with the agency or accessing federal services. Furthermore, if a third-party AI system is deployed, agencies must prohibit the combining or cross-referencing of data from individual interactions with that AI system with any commercial or personal accounts the individual holds with the same vendor.
Defines "substantially influence" as serving as a principal basis for a critical decision, including by producing an approval, rejection, score, classification, or recommendation that an agency uses in whole or in part during an augmented critical decision process.
Section 2(10)
Why it matters:This is standard legislative practice to define terms upfront, but the specific phrasing here has a far-reaching impact on the bill's applicability.
Case for: Supporters would argue that a clear and broad definition is necessary to establish the threshold for when AI's involvement in a decision triggers the bill's transparency and accountability requirements, ensuring that even partial AI influence is covered and preventing agencies from circumventing the rules.
Case against: Critics might argue that this broad definition could lead to an expansive interpretation, requiring agencies to apply the bill's rigorous requirements to AI systems that have only a minor or indirect influence on a final decision, creating unnecessary administrative burden and potentially slowing down government operations.
Estimated impact: Significantly broadens the scope of AI systems and decisions subject to the bill's requirements, impacting potentially all AI systems that contribute to critical decisions, not just fully automated ones.
States that the Act shall have no force or effect beginning on the date that is 10 years after the date on which the Director issues the transparent automated government guidance.
Section 5
Why it matters:Sunset clauses are often included to ensure periodic review and reauthorization of legislation, or to limit the duration of new authorities, particularly in rapidly evolving fields like AI. It can also be a mechanism to garner broader support by making the legislation temporary.
Case for: Proponents of sunset clauses argue they encourage Congress to revisit and update legislation, ensuring it remains relevant and effective in a rapidly evolving field like AI, and prevents permanent, potentially outdated, regulations from remaining on the books indefinitely.
Case against: Critics might argue that a sunset clause creates uncertainty for agencies and the public, potentially hindering long-term planning and investment in compliant AI systems and infrastructure, as the regulatory framework could disappear or change significantly after 10 years.
Estimated impact: Creates a definite expiration date for the entire Act, requiring future legislative action to continue its provisions and potentially leading to a lapse in AI governance if not reauthorized.