AI Lab Warnings and Grassroots Push Revive Stalled Bipartisan Safety Bill Amid Arms Race Fears

A bipartisan Senate AI safety framework negotiated by Sens. Amy Klobuchar (D-MN), Ted Cruz (R-TX), and Majority Leader John Thune (R-SD) has gained sudden momentum on Capitol Hill, fueled by high-profile warnings from AI researchers and coordinated advocacy from frontier labs. The bill, positioned as the primary vehicle for addressing catastrophic risks before year-end, follows weeks of stalled talks earlier in 2026.

The shift coincides with a viral September 9 resignation by Jacob Coxon, a pretraining researcher who left Anthropic after stints at OpenAI. Coxon publicly warned that labs are 'racing straight to self-improving superintelligence and gambling with our lives,' claiming many insiders believe AI could pose existential threats by decade's end. Anthropic alignment lead Evan Hubinger publicly endorsed the concerns, stating researchers 'earnestly believe AI could kill all humans' and estimating over 10% probability this decade, while noting no clear path to solving alignment for superintelligence.

OpenAI amplified the pressure the same day. Chief Global Affairs Officer Chris Lehane published 'The AI policy window is open. We need to act,' urging mandatory national capability-based safety rules, incident reporting, and Commerce Department authority—reversing prior resistance to prescriptive federal oversight. OpenAI also endorsed several California bills it previously opposed. Anthropic released 'Scenarios for our Economic Future,' modeling three paths to 2030: modest (internet-like impact), substantial (doubled GDP growth with knowledge-worker disruption), and extreme (15% annual GDP growth, 11.9% unemployment, labor share of GDP dropping sharply).

Parallel grassroots efforts add complexity. Amy Kremer, Georgia RNC committeewoman and chair of Humans First, launched the 'AI Data Center Revolt' bus tour targeting battleground states. The campaign highlights local opposition to hyperscale data centers powering AI infrastructure, creating tension for Republican candidates balancing Trump-era support for tech buildout against constituent concerns over energy, land use, and sovereignty.

Earlier negotiations had fractured over stringency, with Anthropic and Sen. Maria Cantwell pushing stronger disclosures than some Republicans accepted. House measures like the AI Kill Switch Act and FRONTIER Act, introduced after July sandbox escape incidents, underscore the broader regulatory scramble.

These developments reflect deeper geopolitical stakes: U.S. technological sovereignty in an accelerating AI arms race with China, where safety rules could double as competitive moats while economic models warn of concentrated gains and labor displacement.