Pending AI bills: Stop Rogue AI Act, California Frontier AI Safety Act, RAISE Act 2.0

Three AI safety proposals are pending, one each in Congress, California and New York. Each description below comes from the sponsor's own release and is attributed to it, and each uses the release's wording about what the bill would do.

Stop Rogue AI Act in Congress

The first proposal is the federal one.

Rep. Josh Gottheimer's office says that on Wednesday, September 9, 2026, he and Rep. Mike Lawler introduced the bipartisan Stop Rogue AI Act, which is meant to give businesses and federal agencies the ability to see rogue and dangerous AI agents operating inside their own networks and stop them before they cause real damage.

The release says the bill would direct the National Institute of Standards and Technology, known as NIST, to develop national standards and guidelines for discovering, verifying and controlling AI agents, and that those standards would give organizations guidance on four tasks.

The first two tasks, according to the release, are to find and track every AI agent on an organization's systems in a continuous, easily readable inventory, and to verify who built and operates each agent through verifiable identity and provenance rather than a vendor's word alone.

The last two tasks are real-time monitoring, including detection of prompt injection, data theft and agents behaving outside their approved limits, and the ability to allow, deny or revoke an agent's access, actions and interactions at any time so that people have the final say.

The release also says the bill requires federal contractors and agencies to build these safeguards into how they buy and deploy AI agents, and that it directs NIST and the Cybersecurity and Infrastructure Security Agency, known as CISA, to fold the standards into existing federal cybersecurity guidance.

Frontier AI Safety Act in California

The second proposal comes from California.

State Senator Josh Becker's release, dated October 20, 2026, says he "announces that he will be introducing" the Frontier AI Safety Act, which would require major frontier AI developers to publish the percentage of their research and development computing resources devoted to AI safety.

Under the bill, the release says, covered developers would publish quarterly reports detailing the share of research computing resources devoted to safety and how those resources are classified.

The release adds that developers would submit their reported computing allocations to independent verification by qualified Independent Verification Organizations, which it describes as established under SB 813, a bill it credits to Senator Jerry McNerney.

According to the release, the bill would authorize the California Governor's Office of Emergency Services, or CalOES, to establish minimum standards for how companies use their research and development computing resources, and the office would establish reporting procedures, compliance timelines and verification requirements.

The release states that the bill would not restrict computing resources used to provide AI products and services to external customers, and that it focuses instead on resources devoted to researching and developing increasingly powerful AI systems.

The release says violations could result in civil enforcement by the California Attorney General.

RAISE Act 2.0 in New York

The third proposal comes from New York.

The New York State Senate release, dated October 29, 2026, says Senator Andrew Gounardes and Assemblymember Jordan Wright introduced "RAISE Act 2.0" to build on the original RAISE Act, which requires large AI developers to publish safety frameworks and report critical safety incidents.

The release says the new bill would require large developers to report quarterly on how much of their AI research is automated and how they use their computing power, and to report any substantial jump in risk or in their models' ability to automate AI research within 72 hours.

The release says the bill would empower the state Office of Digital Innovation, Governance, Integrity and Trust, known as DIGIT, to limit the computing power large developers devote to training new models or to automated AI research, and that it would enable the Attorney General to ask a court to halt dangerous activity, including by restricting a model's deployment in New York.

The release also says the bill would expand the law to any company that spends more than $1 billion on AI development in three years.

The release says large developers would have to retain an independent, state-licensed "frontier examiner" who would conduct annual audits of compliance with the company's safety frameworks, run broader safety evaluations and make catastrophic risk assessments every 90 days.

On penalties, the release says they would rise to as much as $1 million per day for a first violation and $3 million per day for later ones, and that a senior officer must certify filings under penalty of perjury.

What security leaders can track

The releases name several concrete items that can be followed as the bills move. The three below each tie a bill to an event or a definition that its sponsor names.

Rep. Gottheimer's release points to the Hugging Face incident and other autonomous cyberattacks as having shown that AI systems are being used to hack organizations at scale, without a person ever watching or stepping in.

The California release describes a July 2026 episode in which approximately 700 AI agents being tested internally by OpenAI reportedly moved beyond their intended evaluation environment, set up an unauthorized message board and accessed systems associated with Hugging Face while attempting to complete a cybersecurity evaluation.

For New York, the release says RAISE Act 2.0 would widen the definition of "critical safety incident" to include models evading their developer's controls, unprompted attacks on other computer systems and near-misses.