Picture a Medicare Advantage plan that uses an AI system to review prior authorization requests for post-acute care. The model flags a claim as low-priority, the denial goes out, and the patient never appeals -- because, as one industry analysis found, three in four health plans now use AI in prior authorization, with appeal overturn rates above 80 percent in Medicare Advantage and patient appeal rates below 1 percent, meaning most incorrect denials are never contested. Nobody broke a law, exactly. The algorithm just optimized. That is precisely the gray zone that Rep. Sara Jacobs (D-CA) is trying to close. On June 3, 2026, Jacobs introduced landmark, comprehensive legislation to hold companies accountable when AI is used to violate existing federal laws. The bill, H.R. 9125, is called the Sectoral AI Governance Act of 2026, and as of this week it is attracting the kind of cross-aisle attention that makes Washington watchers sit up: AI Governance Weekly is tracking whether bipartisan alignment between Speaker Johnson and Minority Leader Jeffries produces a formal bill under the Sectoral AI Governance Act of 2026, which would trigger mandatory algorithmic accountability obligations across multiple regulated sectors. That is a significant signal. The Jacobs bill was a Democrat-only introduction in June; the possibility of leadership-level buy-in from both parties transforms it into something much harder to dismiss.
The core idea is deceptively simple. AI and other algorithmic decision-making systems are increasingly shaping outcomes in people's lives -- whether people get a job, a loan, housing, or health care. Federal agencies enforce laws in all of these areas, but they don't always have a clear, coordinated basis for writing rules when AI contributes to harms that are already illegal. The Sectoral AI Governance Act would give federal agencies a consistent framework for writing and issuing rules whenever the use of AI is likely to materially contribute to violations of existing federal laws. In plainer terms: the bill does not invent new prohibitions. It tells every agency that already has enforcement authority -- the Consumer Financial Protection Bureau, the Department of Health and Human Services, the Equal Employment Opportunity Commission, the Department of Justice -- that it now has a clear, consistent mandate to write rules specifically addressing the AI systems operating inside its domain. The bill's definition of an "algorithmic decision-making system" covers any computational process, including one based on statistics, machine learning, or artificial intelligence, that assists in making or executing a decision, if such process is capable of altering the outcome of the decision, which is broad enough to sweep in credit-scoring engines, clinical-risk stratification tools, and pretrial risk assessments alike.
For hospitals and health systems, the practical stakes are concrete and immediate. CMS already embedded AI into federal health programs in January 2026 when it launched the WISeR model, deploying AI-powered prior authorization review in six states: New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington. The Sectoral AI Governance Act would give HHS the statutory footing to write rules requiring that such systems meet fairness and transparency standards before deployment -- not after lawsuits. For banks, the mechanism is equally direct: a CFPB empowered under the act could require that mortgage-underwriting or credit-card limit algorithms be tested for disparate impact against protected classes before they go live, rather than waiting for enforcement actions case by case. If a rental screening algorithm trained on biased data downgraded applicants from majority-Black or Hispanic zip codes, even when their income and credit are identical to those of approved applicants, that could violate the Fair Housing Act; the bill would allow the Department of Housing and Urban Development to require safeguards such as testing rental screening tools for discriminatory patterns before they are used. The pattern is the same across every regulated sector: pre-deployment accountability, not post-harm litigation.
The procedural architecture of the bill is as important as its ambitions. Not later than 60 days prior to issuing a notice of proposed rulemaking, the head of an agency shall publish an advanced notice of proposed rulemaking to solicit public comment on whether the use of the algorithmic decision-making system that the agency proposes to regulate is likely to materially contribute to violations of the federal law that the agency is authorized to enforce. And the agency head must consider whether the use of the algorithmic decision-making system occurs in connection with the administration of a government service or public benefit, including by a contractor or service provider, and, where practicable, seek to mitigate unnecessary adverse effects of the rule on the delivery, accessibility, timeliness, integrity, or continuity of such service or benefit. That contractor clause matters enormously: it means that a hospital cannot hand its utilization-review algorithm to an outside vendor and consider itself off the hook. A federal court recently allowed a lawsuit to proceed against a company that relied on a vendor's AI tool for consequential decisions, finding the deploying organization could be held responsible for the allegedly discriminatory outputs that the tool produced. The message for boards is direct: choosing a vendor does not transfer your organization's legal accountability for what that vendor's tool does inside your workflows. The bill would codify that logic government-wide.
Skeptics have real ammunition. At the federal level, focus has largely been on governing frontier model risks, but no bills have passed in the current session; an unfortunate trend in 2026 is the lack of momentum around regulation of the use of automated decision systems in consequential decisions. Rep. Jacobs herself has been blunt about the obstruction: since January of last year, Speaker Johnson cancelled 62 of 261 scheduled voting days -- almost a quarter of the days that Congress could have been in session to pass any one of the bipartisan AI bills sitting in the chamber, including 62 days that could have moved the 89 recommendations the Bipartisan AI Task Force identified almost two years ago. Innovation-industry lobbyists argue that empowering dozens of agencies simultaneously could produce contradictory rules that paralyze AI deployment in sectors where it is already saving lives and reducing costs. That concern is not trivial; the bill addresses it by directing the Director of the Office of Management and Budget, acting through the Office of Information and Regulatory Affairs and in consultation with the Director of the Office of Science and Technology Policy, to issue guidance to resolve conflicts and ensure consistency across agencies. Whether OMB can actually coordinate a dozen powerful regulators writing AI rules in parallel is a question the bill's text cannot fully answer. What the bill does make undeniable is the baseline: federal laws shouldn't become optional just because technology is new; AI is already helping make life-altering decisions for millions of Americans -- whether they get a loan, a job, or health care coverage -- but too often, it's operating in a gray area.
If bipartisan leadership pressure converts H.R. 9125 into law, the United States will have done something the EU's top-down AI Act did not: built algorithmic accountability directly into the enforcement machinery of every regulator that already has power over American life, which means the fight over AI governance moves immediately from Capitol Hill to the rulemaking dockets of roughly 70 federal agencies.