Federal AI bills would add notice and human review, create a CISA task force

The bill would also call for redress processes, including human review, for people who receive adverse critical decisions made using agency AI systems. That could give someone a way to ask a person to review a decision; it would not guarantee that the decision is reversed.

The proposal addresses a visibility and recourse problem: people may not know when AI is involved in a federal interaction or consequential decision, or how to challenge the result. Peters has said people should know when federal AI is involved and be able to seek human review. Lankford supports transparency and oversight while citing security, privacy and unintended consequences as concerns.

The accounts available here do not answer which systems or decisions would be covered, what notices would have to say or how agencies would operate the review process. Those details matter in practice. A notice framework depends on agencies identifying the systems in use and making the information useful to the people affected.

The CISA bill would coordinate risk work, not regulate AI broadly

Reps. Troy Carter and Bennie Thompson reintroduced the CISA Securing AI Task Force Act on Thursday, Oct. 8. The proposal calls for a task force to be established within one year of enactment. Its duties would include coordinating agency work on safe and secure AI, assessing relevant initiatives, advising stakeholders on cyber risks and identifying workforce gaps. It would also require briefings to House and Senate homeland-security committees twice a year.

Thompson, the House Homeland Security Committee’s ranking member, said in a statement reported by Nextgov/FCW: “This bill creates a dedicated CISA task force so the agency can better understand the risks that come with AI and make sure we are prepared to protect the systems and infrastructure Americans rely on every day”. That is the sponsors’ rationale, not evidence that the task force would reduce risk.

As described, the task force would have an internal coordination and advisory role, not act as a general AI regulator or independent testing body. The roundup does not establish what staffing, resources or authority it would receive. Without those details, the proposal’s capacity to coordinate agencies or address workforce gaps cannot yet be assessed.

Both proposals depend on agencies’ ability to identify AI use

The TAG Act would pair notice with a potential route to human review. The CISA proposal would organize government expertise on AI-related cybersecurity and safety risks. Neither, as described, establishes independent checks on every agency system.

The roundup says the latest federal inventory reported more than 3,600 agency AI uses in 2025. It does not establish whether that inventory is complete. The figure indicates the reported scale of government use, but a notice framework would still depend on agencies identifying systems consistently and explaining their role in decisions to affected people.

Both measures remain proposals, and the roundup does not report that agencies have begun implementing them. Whether either bill advances, and what lawmakers specify about coverage, review procedures, task-force authority and resources, will determine what the proposals can deliver.