Altman Accepts AI Harms for Benefits, Tests Washington and Florida Models
OpenAI chief executive Sam Altman said the world should accept some bad things happening with AI in exchange for the benefits of the technology.
Altman cited hacks, scams, misuse and other bad things as examples that will happen with AI. He advocates a lighter-touch regulatory stance that accepts some bad things will happen as society figures out resilience, according to The Guardian.
Reuters quoted the formulation as: "We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency..." in its account of the interview. The Guardian attributes the remarks to an interview for Politico's Decoded podcast released on Monday. Politico published the original interview in which Altman made the case for accepting risks for benefits, as documented in the Decoded interview.
The comments followed a pact at the White House last week signed by Donald Trump and AI bosses including Altman and Dario Amodei of Anthropic. Under that arrangement, Trump said companies should self-police to limit risks from AI-enabled cyber-attacks and bioweapons.
The response from Florida was adversarial. Ron DeSantis, governor of Florida, responded to the interview saying he had "no dice" with a handful of tech oligarchs deciding safety for others. The State of Florida asked a judge to bar OpenAI from developing new AI models without third-party approved guardrails.
Altman’s risk tolerance framing also coincided with internal dissent and public anxiety. An OpenAI safety leader quit the company, warning that the company’s culture was broken and that AI firms were not being nearly careful enough, according to The Guardian’s Oct. 3 report. A separate Reuters report on the resignation framed the warning as a claim that the time for trial and error was over. Polling underscores the political exposure. Three-quarters of Americans worry that AI companies have not gone far enough to prevent AI from causing serious harm to society, Reuters reported.
Altman’s current position is consistent with his prior resistance to pre-release controls. He planned to urge U.S. lawmakers against requiring AI developers to obtain government approval before releasing new models, Reuters reported in June. On corporate and safety thresholds, Altman said OpenAI will not hold an IPO in 2026 and feels no pressure to go public. He said a 10% risk of AI causing extinction is "unacceptable," according to Reuters reporting from September.
The broader context here is a collision of regulatory theories. Washington is testing voluntary, company-led policing for catastrophic-use categories. Florida is testing judicially enforced, third-party approved guardrails as a precondition for development. Congress is being asked to decide whether pre-release government approval should exist at all. The split is jurisdictional. It is also philosophical, between resilience after deployment and constraint before it.
Looking at what this means for diplomacy and governance, the credibility test will be empirical. Altman’s argument assumes harms such as fraud and intrusions can be absorbed while defenses mature. Critics, including state-level challengers and departing safety staff, assume the trial-and-error window has narrowed. Public opinion data suggests elected officials have incentives to press for harder guarantees. The next markers to watch are judicial: whether a court entertains a development freeze tied to external review, and legislative: whether self-policing pledges acquire enforceable standards or remain voluntary commitments.
