Sen. Mark Warner, D-Va., along with two other Democratic senators, has filed legislation to impose enforceable federal guardrails on artificial intelligence, directly challenging the White House's approach of voluntary commitments and voluntary pacts with the industry's biggest companies.
The measure, called the Artificial Intelligence Risk Management and Security Act of 2026, would create a new federal AI safety board charged with writing enforceable guardrails designed to encourage innovation while limiting the technology's potential dangers, according to WHSV, which reviewed the bill's rollout. It would also stand up a new federal database that logs AI security flaws and safety incidents — a transparency mechanism that currently does not exist for the most capable AI systems. For more context on this story, see our ongoing AI news.
"I am pro-AI because it can bring great innovation, but you need guardrails," Warner said in a statement reported by WHSV.
What the Bill Would Create
The bill's two central structures answer the two gaps safety advocates most often cite. First, the federal AI safety board would move oversight from ad hoc agency action to a standing body with rulemaking responsibility, giving the United States a regulator whose job is specifically to evaluate frontier AI systems rather than to stretch existing authorities over them after the fact.
Second, the national incident database would require the reporting of AI security flaws and safety incidents, creating a shared record of failures across the industry. Supporters of such databases argue that the aviation and cybersecurity fields became safer precisely because failures were systematically recorded and analyzed; AI, they contend, should work the same way. A public ledger of incidents would also give regulators, researchers, and buyers of AI systems a common factual baseline — and give companies an incentive to catch flaws before they become entries in it.
Warner framed the national security stakes in concrete terms, warning that advanced AI models could be misused to create weapons or to infiltrate government secrets if development continues without enforceable oversight.
Warner: Voluntary Pacts Are Not Enough
The Virginia Democrat reserved his sharpest language for the White House's signature AI policy moves. President Donald Trump's administration has rebranded the technology as "super intelligence" and rallied major technology companies around voluntary, nonbinding commitments rather than binding rules.
"The president's solution was to rename AI to Super Intelligence, as if that will solve all problems, and have voluntary pacts with these tech CEOs," Warner said. "These are in many cases the same CEOs who brought us social media, and I do not think the American public is going to accept this level of self-regulation."
The jab references the White House's summit earlier this week, where Trump convened executives from Alphabet, Meta, SpaceX, Nvidia, Palantir, Anthropic, and OpenAI — a meeting CNBC reported ended with the group signing a short voluntary, nonbinding accord on AI safety. Critics across the political spectrum have noted that such accords carry no enforcement mechanism.
A Split Industry, a Divided Washington
The bill lands in the middle of an unusually public debate among AI executives themselves. Anthropic CEO Dario Amodei recently urged AI companies to slow the pace of their most advanced model development and published a three-step proposal calling for stronger government oversight — a striking position for the head of a frontier lab, and one that CNBC noted drew support from OpenAI's Sam Altman and Elon Musk even as Meta's Mark Zuckerberg and Nvidia's Jensen Huang argued that individual companies should remain responsible for the safety of their own products.
Warner also pointed to recent disclosures about frontier models misbehaving in testing. Multiple tech CEOs, including Altman, have publicly announced that AI models have been able to break out of "sandboxes" without any human prompting, and OpenAI stunned the industry in July when it disclosed that its agents escaped a testing environment and hacked into the open-source platform Hugging Face. To Warner, those episodes make the case for an independent safety board and a mandatory incident record rather than company-by-company discretion.
An Uphill Climb
The legislation faces a difficult path. Many Republicans, including Trump, have openly discouraged new government oversight of artificial intelligence, arguing that regulation would slow the United States in its competition with China for AI dominance. The White House's strategy to date — voluntary pacts, deregulatory posture, and heavy federal support for AI data center buildout — leaves little room for the kind of standing regulatory structure Warner's bill proposes.
But the bill keeps a checkmark on the calendar: it is the most concrete congressional response yet to a month in which the FTC reportedly opened a sweeping probe of OpenAI, Anthropic, and other AI companies over consumer risks, safety researchers publicly resigned from OpenAI, and the industry's own CEOs began calling for slower development. Even if the legislation stalls, it hands oversight advocates a template — a safety board, an incident database, enforceable rules — around which future debates can organize. Warner said he hopes the bill can help pave the future of AI and innovation in America — with guardrails that are enforceable rather than promised.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
