Sam Altman says world must accept some harm from AI development

OpenAI chief executive Sam Altman has argued that society must accept a degree of inevitable harm as the necessary cost of giving people access to artificial intelligence, drawing a sharp philosophical line against rival developers who urge caution.

Speaking in an interview published in Politico's technology newsletter Decoded, Sam Altman stated that demanding an ecosystem entirely free of technological abuse would require an intolerable level of centralized control that stifles human agency. According to reporting by Shattered, the OpenAI leader explicitly declined to trade the technology's widespread gains for a guarantee of zero hacks, zero scams, or zero misuse.

Media additions

"We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency,"

Sam Altman, OpenAI CEO, via Politico

As noted by Devdiscourse and The Daily Star, Altman told Politico that the positive economic, scientific, and personal utility generated by public access will outweigh downstream harms by orders of magnitude because people will ultimately do far more good than bad with the technology.

This pragmatic stance establishes a public philosophical divide with rival frontier lab Anthropic. Altman noted that while he understands the perspective that AI is so dangerous that a single laboratory in San Francisco should hold it and dole out the benefits, he considers that approach a completely unacceptable trade-off that runs counter to OpenAI's preferred lighter-touch regulatory stance.

In response to these remarks, an Anthropic spokesperson clarified to Politico that its proposed containment measures apply exclusively to frontier models possessing extreme operational capabilities rather than routine consumer software or standard utilities. The divergence follows an essay published in September by Anthropic chief executive Dario Amodei, who urged the industry to deliberately slow down capability releases to allow safety evaluations and containment frameworks to mature. Although Altman initially expressed rhetorical sympathy with Amodei's call for pacing the frontier, his latest interview demonstrates a fundamental strategic split.

Tensions within the sector have intensified alongside internal safety concerns. According to The Daily Star and Devdiscourse, Anthropic researcher Jacob Coxon resigned in September, stating that those building artificial intelligence believe it could pose severe terminal risks by the end of the decade.

Despite these warnings, Sam Altman drew a firm line against what he categorized as existential catastrophes. He carved out a strict exception for a potential catastrophic outcome, distinguishing recoverable everyday frictions—such as localized hacks and financial scams—from irreversible long-term dangers.

Altman's posture aligns closely with the direction of federal policy in Washington. US President Donald Trump has consistently dismissed calls for heavy compliance mandates, arguing that restrictions would hamper domestic firms competing with international rivals in China. The administration maintains that existing legal frameworks enforced by the Department of Justice and other regulatory agencies are sufficient to prosecute illicit actors.

The debate over acceptable harm arrives as industry leaders navigate shifting oversight frameworks. Meanwhile, OpenAI is scheduled to testify under oath at a New York City Council hearing on artificial intelligence safety alongside other major AI firms.

The coming months will test whether regulators, lawmakers, and federal task forces accept Altman's premise that everyday misuse is a tolerable cost of technological progress, or whether mounting pressure from civil liberties groups, safety advocates, and officials will force a shift in corporate liability standards.