Core Event: Tech Executives and Politicians Clash Over “Pacing” AI Development
Last Saturday, Dario Amodei, co-founder and CEO of Anthropic, published a long-form essay titled “We Must Pace the Frontier”, calling for a deliberate slowdown in artificial intelligence development to prioritize safety. The essay triggered a wave of responses from tech leaders, policymakers, and public figures across the ideological spectrum.
- Release date: Saturday, September 13, 2026
- Core proposal: “Pacing” — slowing down but not stopping progress; implementing embedded third-party evaluators; coordinating standards among democratic frontier AI companies; seeking global coordination, “to the extent possible”, between democratic and authoritarian governments
- Anthropic’s commitment: Unilateral adoption of the first step — embedding independent third-party evaluators
- Definition clarified: “Pacing” means companies take adequate time to align and safeguard models, with third-party verification, not halting training or technical progress
Polarity of Responses: Support, Skepticism, and Outright Rejection
The supporting camp includes OpenAI CEO Sam Altman, who said “I agree with Dario that we need to pace the frontier” and committed to independent evaluators at OpenAI; Google DeepMind chair Demis Hassabis called the essay “pointing toward the right path forward”; former U.S. Vice President Kamala Harris declared that “we must responsibly slow the pace of frontier AI development” to ensure AI serves the public interest; and former Transportation Secretary Pete Buttigieg called for safeguards and a “kill switch on rogue AI.”
Skeptical voices—notably Vice President JD Vance—questioned the motivation behind companies “begging the government to regulate them,” suggesting it may be a “Trojan horse.” Vance also cited concerns about China’s long-term AI ambitions.
Opposition came more forcefully from former President Donald Trump, who tweeted that the only “control or guardrails” AI needs is “a STRONG AND SMART (High IQ!) PRESIDENT,” asserting that his administration has already stopped AI “people” from doing bad things and dismissing the current safety movement as a “SICK conspiracy.” NVIDIA CEO Jensen Huang, speaking onstage after a mock phone call with Trump, stated “the robots will not be taking over” and emphasized inclusive American AI leadership.
Tensions Between Industry self-regulation and Government oversight
A key irony is that the most advanced frontier AI labs are voluntarily proposing tighter constraints, while legislators express suspicion about corporate motives. House Speaker Mike Johnson said he would convene the companies “tomorrow” to discuss the “right balance,” but stressed that a moratorium is impossible because “China will overtake us”—highlighting the core conflict between safety and national competitiveness.
Former FTC Chair Lina Khan reminded the public that existing laws already cover dangerous products, and the FTC repeatedly stressed during her tenure that “there’s no AI exemption from laws already on the books.” Former White House AI czar David Sacks added that the push to slow development need not be purely altruistic: “stop pretending you need anyone else’s permission.”
Practical Takeaways for Different Stakeholders
- For AI developers and startups: Third-party evaluation is becoming an industry expectation; early preparation for compliance and engagement with evaluation frameworks will be competitive advantages;
- For policy watchers in the U.S.: Congress has not yet introduced formal AI legislation, but agencies like the FTC are signaling they can act under current authority—suggesting incremental enforcement rather than sweeping new laws;
- For end users and investors: Open-source release of high-capability models will likely remain limited until safety is verified publicly; patience for rigorously validated, audited releases is advisable.
In closing
The debate is not about whether AI should be safe, but about who defines safety and determines when “enough time” has passed. As frontier labs advocate for self-imposed constraints while facing widespread political skepticism, AI governance has entered a new, more contentious phase defined by strategic competition rather than consensus—and as Amodei notes, global coordination must proceed “to the extent possible” given real-world political constraints.