In an interview on October 04 with Politico’s Decoded, Sam Altman candidly pointed out a “huge daylight” between OpenAI and Anthropic regarding the extent of government intervention in frontier AI development. His response was notably more direct and pragmatic than the usual generic statements.

Specifically, Altman affirmed: “We believe the world should accept some bad things happening in exchange for the benefits of this technology and giving people agency.” He argued that centralizing oversight authority in “a single lab in San Francisco” to completely prevent risk is an exchange that is “totally unacceptable.” At the same time, he made clear his position that he would not make this trade-off merely to provide absolute protection against major hacks, tech abuse, or scams.

The paradoxical similarity between both sides

The context highlighted in the interview offers an especially noteworthy detail. Although Anthropic has generally been viewed as the strongest proponent of AI regulation, while OpenAI has opted for a softer approach, Politico noted that the gap between the two organizations has narrowed significantly recently. Nevertheless, a “fundamental philosophical gap” remains between them.

This convergence is clearly demonstrated through specific moves:

  • Altman expressed agreement with Dario Amodei’s proposal last month to slow the pace of frontier model development.

  • OpenAI proactively followed Anthropic’s lead in supporting stricter state-level safety regulations, marking a shift from its earlier direction.

  • OpenAI’s policy advocacy team recently backed a bipartisan bill in the House of Representatives requiring leading AI companies to undergo third-party independent safety evaluations.

Thus, the “daylight” Altman referred to does not stem from these specific regulations at all. He clearly distinguished between two aspects: on the one hand, compliance and cooperation with legal rules once enacted, and on the other, a principled divergence regarding the degree of risk one is willing to accept.

What is the real gap?

Notably in the interview, when discussing the limits of government intervention, Altman focused on the principle of risk tolerance rather than providing specific figures or thresholds. In the U.S., this remains a core subject sparking endless debate.

While President Trump signed an executive order changing the term “AI” to “Super Intelligence,” a task force led by Director of National Intelligence Jay Clayton also recently released a charter with a 120-day deadline to complete a report. Their mandate is to evaluate risk levels while ensuring they “prevent over-regulation and regulatory capture.” Against this backdrop, Altman’s perspective, though public, is not yet accompanied by any detailed enforcement mechanisms.

The above analyses merely reflect the current state of the policy debate rather than serving as forecasts. Future developments will not depend on the statements of an individual corporate leader, but will be determined by the report in the next 120 days as well as the ability to balance safety requirements and anti-capture provisions in forthcoming legislation.