OpenAI CEO Sam Altman has reignited the debate surrounding AI regulation by suggesting that a certain level of risk must be accepted to reap the benefits of artificial intelligence.

His argument is that rather than attempting to completely block AI misuse or accidents, society should tolerate some side effects that arise during the technological advancement process. This stance contrasts with the "AI speed-regulation theory" championed by Dario Amodei, CEO of rival firm Anthropic. While Altman has previously agreed in principle on the need to secure AI safety following Amodei's remarks on speed control, this latest statement reveals a clear divergence in perspective.

In an interview on the 4th (local time) with Politico’s tech podcast Decoded, Altman stated that there are significant differences in the regulatory views of OpenAI and Anthropic. He remarked, "I think you have to accept some bad things in order to get the benefits of the technology and keep human agency."

He stated that he disagrees with the notion that because AI could become overly powerful and dangerous, only a single research lab (AI development company) should hold powerful AI technology while figuring out ways to prevent side effects and distribute benefits. He described OpenAI's stance—which favors regulation itself but opposes excessively heavy regulation—as a "light-touch regulatory stance."

Among Altman's remarks, the part most likely to stir controversy is his assertion that negative outcomes caused by AI must be accepted to a certain degree.

He explained that an approach such as "ensuring that no major hacking incidents occur, preventing AI technology from being misused, and not allowing a single fraud case" is difficult to accept. This is because he believes people will utilize AI for good things far more often—dozens of times more—than for bad things.

However, Altman drew the line by clarifying that this does not mean accepting all risks. He noted that he does not mean enduring "catastrophic risks," such as humanity seriously losing control over AI.

This statement comes amid a recent intensification of the regulatory debate between OpenAI and Anthropic. While OpenAI has traditionally maintained a relatively cautious stance on AI regulation, recent events such as external hacking incidents targeting its AI models have led it to partially align with the AI safety and speed-regulation discussions raised by Anthropic.

Even so, Altman's latest remarks demonstrate that while the two companies have found some common ground on the necessity of AI safety, a significant gap remains on the core question of how much risk should be tolerated while continuing AI development.

In particular, Altman mentioned the controversy surrounding "regulatory capture" raised in connection with Amodei's regulatory proposals. This criticism suggests that regulated companies might demand regulations favorable to themselves to make it difficult for competitors to enter the market or catch up.

In response, Anthropic has countered that its regulatory proposals apply only to cutting-edge, front-runner AI models and thus actually benefit latecomer competitors, making the claim of regulatory capture to block competition inaccurate, according to Politico.

Consequently, a social consensus is needed on just how far the "level of risk" raised by Altman extends. Altman has sparked yet another key point of contention by making it clear that society must endure a certain level of failure and misuse for the immense benefits AI will bring.