Full Breakdown
AI Safety Debate Intensifies as Anthropic Calls for Government Oversight
By Drooid · · How we work
Core Event: Anthropic’s Policy Chief Says Honor-Code Self-Regulation Won’t Work
On September 16, 2026, Sarah Heck, Anthropic’s head of public policy, told attendees at POLITICO’s Decoded Summit in Washington, D.C., that artificial-intelligence firms cannot be trusted to police themselves with an “honor code.”
Background & Context: Industry Push for Slower Development
The remarks came days after Anthropic CEO Dario Amodei published an essay titled “We Must Pace the Frontier” (dated September 12, 2026). The essay proposes a three-step plan—embedding third-party evaluators, coordinating safety standards among democratic frontier AI firms, and seeking limited cooperation with authoritarian states. Amodei’s call was quickly endorsed by OpenAI CEO Sam Altman and SpaceX AI chief Elon Musk.
Resignations have amplified concerns. Former Anthropic researcher Jacob Coxon left the company, warning that AI labs are “gambling with our lives.” Around the same time, former Google DeepMind researcher Bilal Chughtai resigned and joined the nonprofit BlueDot Impact, citing the possibility that AI could “kill us all.”
Official Statements & Responses
- Anthropic: Heck said the company is in daily contact with the White House and has been working with Congress “all year,” seeking concrete regulatory frameworks rather than voluntary codes.
- He also reaffirmed Anthropic’s commitment to embed an external review team with internal-level access.
- OpenAI: Lobbyist Chris Lehane confirmed OpenAI’s support for the bipartisan House provision that would require top AI firms to embed outside evaluators.
- U.S. Government: President Donald Trump rebuked Amodei on September 14, 2026, dismissing additional AI guardrails as a “HOAX” and asserting that only a “strong and smart” president can provide control.
Data & Statistics
- Two high-profile resignations (Jacob Coxon, Anthropic; Bilal Chughtai, Google DeepMind) have been publicly linked to safety concerns.
- The FRONTIER Act, co-authored by Rep. Jay Obernolte (R-CA) and Rep. Lori Trahan (D-MA), proposes mandatory third-party evaluators for frontier AI models.
Conflicting Reports & Gaps
Sources differ on the preferred regulatory path. Anthropic and OpenAI lobbyists advocate for statutory third-party oversight, while the Trump administration favors limited, industry-led safeguards, arguing broader regulation could cede advantage to China. No consensus exists on whether voluntary safety standards can adequately mitigate existential risk.
Verbatim Quotes
- “We want to work with government to figure out what makes sense,” — Sarah Heck, former commerce secretary
- “The United States needs to stay in the lead on AI, and you can’t do safety from second place,” — Sarah Heck, former commerce secretary
- “I earnestly believe that AI has the potential to kill us all, and that we might be running out of time to avoid this outcome,” — Bilal Chughtai
- “I think that’s going to be very difficult because the incentives to pull ahead and the military advantage that you get from that are so large,” — Dario Amodei, after anthropic CEO
- “The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT,” — Donald Trump, made by president
What’s Next
Anthropic is continuing its “daily” discussions with the White House and congressional staff, while lawmakers are reviewing the FRONTIER Act’s evaluator provision. The outcome of these negotiations will shape whether the U.S. adopts statutory oversight or relies on industry self-regulation as AI capabilities accelerate.