OpenAI, Anthropic Want to ‘Slow Down’ AI? David Sacks Says They Can Do It Themselves — ‘Stop Pretending You Need Anyone Else’s Permission’
Former White House AI and Crypto Czar David Sacks argued Saturday that OpenAI and Anthropic don’t need lawmakers or regulators to "pace the frontier" if they truly believe their next models are too risky to release. He made the point as questions swirl around Jacob Coxon’s resignation and viral warning about self-improving AI, a flare-up that has fueled public suspicion about who benefits when labs call for slower progress.
In a post on X, Sacks wrote that if Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman want to slow down, they should do it themselves because they already sit at the cutting edge by common measures such as adoption, growth and model strength. He framed their position as a de facto two-player market at the top end of model capability, arguing they are effectively the ones defining what counts as the frontier.
Should AI Developers Self-Regulate For Safety?
Sacks’ message leaned on a simple scenario: the labs insisting on restraint are the same ones with the most control over whether the most capable systems ship. He said that if internal work is alarming enough to justify delay, he supports choosing caution, but he rejects the idea that outside "permission" is required.
The call to action from Sacks comes amid a backdrop of heightened concerns about AI safety and alignment. His call to action comes amid heightened concerns about AI safety and alignment. Altman recently said the company would not go public in 2026, amid broader discussions about AI safety, alignment and the risks posed by increasingly capable models.
Additionally, Evan Hubinger, a researcher at Anthropic, highlighted the existential risks AI could pose, estimating a greater than 10% chance of AI posing a threat to humanity within a decade. These concerns are echoed by former Anthropic researcher Jacob Coxon, who has been vocal about the potential dangers of AI, despite criticism from Sacks, who challenged the evidence behind Coxon’s claims.
Sacks also flagged the speed of Coxon’s sudden reach, pointing to the account’s apparent lack of prior posts or followers before the resignation thread took off. He said that if earlier activity existed, it appeared to have been removed, and he argued the rapid traction made the rollout look coordinated rather than organic.
On the same podcast discussion, co-host Jason Calacanis said Coxon had contacted the show seeking an appearance and then canceled the morning the interview was scheduled. Sacks suggested the pullback may have come after Coxon realized "it would be a tough interview," and he said he wanted to ask who connected Coxon with the Wall Street Journal before the post circulated.
The High Stakes Of AI Liability Risks
In his X post, Sacks tied the push to slow down to downside risk that looks less like altruism and more like legal exposure. He argued that if a model enables a major cyber incident, the companies behind it could face enormous product-liability consequences, and he said markets already penalize systems that act in unexpected or unauthorized ways.
That liability theme also surfaced in the podcast debate around Anthropic’s reported IPO ambitions, which were described as targeting a $2 trillion valuation. The panel argued that if internal safety personnel believe the technology could pose existential danger, that belief could create disclosure complications for Securities and Exchange Commission filings alongside potential product-liability claims.
Anthropic alignment lead Evan Hubinger publicly supported Coxon and said he sees greater than a 10% chance that AI causes human extinction within a decade. The comment added fuel to scrutiny over whether dire internal assessments are consistent with continued scaling and commercialization.
Sacks’ X post also described a business rationale for trading maximum capability for steadier behavior, arguing that customers value reliability and predictability. He portrayed "alignment" as potentially overlapping with ordinary product expectations rather than a separate moral mission.
Check out more of Benzinga’s Consumer Tech coverage by following this link.
Read Next: California Should Allow Sale of Chinese EVs Amid High Gas Prices, Says Ripple Co-Founder
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
Photo courtesy: Shutterstock