New AI safety center has no teeth—companies can ignore its standards

What it does

This bill establishes a new Center for AI Security and Innovation within the National Institute of Standards and Technology (NIST) to evaluate risks from artificial intelligence systems, develop voluntary security standards, and assess U.S. competitive position against foreign adversaries. The Center will work with AI developers on a voluntary basis to test frontier AI systems for national security and economic risks, but has no regulatory or enforcement power—it is purely advisory and research-focused.

Why we flagged it

The bill's core mechanism is establishing a non-regulatory research center within NIST to develop voluntary AI security standards and conduct risk assessments. It is explicitly prohibited from having enforcement or rulemaking authority, making it a research and advisory body rather than a regulatory one.

What the text implies

• The confidentiality shield (Section 5304(g)) prevents federal, state, local, or tribal regulators from using information shared with the Center to regulate AI companies—effectively creating a safe harbor for companies that voluntarily participate in Center evaluations.
• The 'voluntary adoption' framework means AI developers can ignore standards and best practices with no consequence; the Center can only assess whether they are adopting them, not compel adoption.
• The definition of 'covered frontier system' is left to the Director to determine within 180 days, creating regulatory ambiguity about which AI systems fall under the Center's purview.
• The Center may conduct classified evaluations of AI systems but has discretion to publish or withhold results, potentially limiting public transparency on national security risks.
• The 5-year sunset clause means the Center must be reauthorized by Congress; if not renewed, institutional knowledge and ongoing risk assessments could be lost.

Section numbers refer to the bill text the analysis read — linked under Primary records below.

Who it affects

The bill creates a public research and advisory body with no enforcement authority, which could improve AI safety transparency and standards-setting—a public benefit. However, the confidentiality shield (information from AI developers cannot be disclosed or used for regulation) and the voluntary-only framework mean companies retain full discretion over whether to participate or adopt standards, potentially limiting real-world safety impact for ordinary users.

Who stands to gain

• AI development companies (voluntary participation in testing and standards-setting without regulator
• Large AI model developers (covered frontier system evaluators gain early access to security guidance

Named in the bill

National Institute of Standards and Technology (NIST), Department of Commerce, Office of Science and Technology Policy, Department of Energy, Department of Defense, Department of Homeland Security, Intelligence Community, AI developers and covered entities

Where it stands

7 cosponsors: 5 Republicans, 2 Democrats.

• Jun 18, 2026 — Introduced · Congress.gov: “Introduced in House”
• Jun 18, 2026 — Referred to House Committee on Science, Space, and Technology · Congress.gov: “Referred to the House Committee on Science, Space, and Technology”
• Jun 25, 2026 — Markup held in committee · Congress.gov: “Committee Consideration and Mark-up Session Held”
• Jun 25, 2026 — Reported out of committee · Congress.gov: “Ordered to be Reported in the Nature of a Substitute (Amended) by the Yeas and Nays: 29 - 0”

Dates and quoted wording are Congress.gov's action record; the timeline shows status changes, not every procedural step.

Money around this bill

2 lobbying clients named this bill on 2 disclosure filings across 1 quarter, Jun 2026 to Jun 2026. Those filings disclosed $840,000 in lobbying spend. A filing names 60 bills on average, so that figure is what each filing reported, not a share belonging to this bill.

More lobbying clients named this bill than 41% of bills with at least one filing.

Jay Obernolte, the sponsor, reported $976,024 in PAC receipts in the 2026 cycle. $5,000 of that came from 1 PAC tied to these lobbying clients.

• AFL-CIO — $760,000 on 1 filing
• Dell Technologies, Inc. — $80,000 on 1 filing

Lobbying Disclosure Act filings through Jul 20, 2026. A filing shows who paid to lobby on a bill it names, not what changed.