# Anthropic to Revise AI Safety Bill Draft Following Industry Backlash The company commits to updating legislative language after criticism regarding state law preemption. - **Source:** [The Next Web](https://brifing.si/source/afe24b20-b275-4c19-b9a6-8bb3c8babae7) - **Published:** 2026-10-20 - **Topic:** [Business](https://brifing.si/c/business) - **Original article:** - **On Brifing:** https://brifing.si/s/1dd62203-4f1e-4501-9f21-1e0ee117e93d ## Legislative Draft Controversy Anthropic is revising a confidential draft AI safety bill currently circulating on Capitol Hill. The document, leaked by Politico, faced immediate backlash from safety advocates and lawmakers. Critics argue the proposal is weaker than existing bipartisan efforts like the FRONTIER Act, potentially hindering the government's ability to block dangerous AI models. ## Strict Risk Thresholds The draft mandates that the Commerce Department can only block a model if risks are foreseeable and imminent. Thresholds for intervention include over $1 billion in damages or significant casualties. These criteria, mirroring California law, require the government to petition federal courts, a process critics claim is overly restrictive and cumbersome. ## Preemption and State Laws A major point of contention involves the bill's attempt to preempt state-level AI regulations. Critics like Senator Scott Wiener argue this would gut essential public health protections. Anthropic’s federal affairs lead, Anthony Cimino, acknowledged the language was not precise enough and promised updates to ensure the federal standard remains robust. ## Evaluator Independence Concerns The proposal allows independent evaluators to register with the Commerce Department, even if paid by the labs they audit. Experts warn this creates a conflict of interest, leading to potential evaluator shopping where labs select the least demanding auditors. Anthropic is now developing new policy proposals to address these specific oversight gaps. ## Comparison with Existing Bills Observers frequently compare the draft to the bipartisan FRONTIER Act introduced in July by Representatives Jay Obernolte and Lori Trahan. Unlike the FRONTIER Act, Anthropic’s version lacks specific licensing standards for auditors. This absence of formal certification is viewed by legal analysts as a significant weakness in the proposed regulatory framework. ## Startups and Market Impact David Sacks, former White House AI advisor, warned that the draft’s training thresholds are too low. He argued this would capture most Silicon Valley startups, imposing heavy compliance burdens on smaller firms. This creates a tension between Anthropic’s desire for safety and the broader industry's need for a viable innovation environment. ## Context of Recent Incidents The legislative push follows a challenging summer for Anthropic, marked by export control restrictions on advanced models. CEO Dario Amodei has publicly advocated for industry deceleration. These incidents, combined with various rogue-AI reports, have intensified the pressure on Congress to finalize comprehensive safety rules that balance innovation with national security. ## Future Policy Adjustments Anthropic maintains that its proposal is intended to be the strongest framework currently on the table. The company plans to refine its stance on state authority and auditor standards. Moving forward, the effectiveness of these revisions will depend on how well they align with broader congressional goals for federal AI oversight. ## Sources covering this story - https://thenextweb.com/news/anthropic-leaked-ai-safety-bill-backlash-frontier — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._