Sam Altman has argued that society should accept some harmful consequences from artificial intelligence rather than pursue a model that seeks to eliminate every form of misuse, saying the technology's benefits and public access justify the trade-off.
The OpenAI chief executive made the comments in an interview, where he described a clear difference between his company and rival Anthropic over how AI should be governed.
Altman also rejected a regulatory trade-off in which preventing every major hack, misuse or scam meant concentrating control of powerful AI in the hands of a single laboratory.
He drew a firm boundary around what he considers unacceptable, saying catastrophic risks such as a serious loss of control over AI should not simply be tolerated. The comments come as OpenAI has recently moved towards supporting additional AI safety requirements, although Altman continues to favour a lighter-touch approach than Anthropic.
Sam Altman Draws the Line on AI Risks
Altman was asked about the difference between OpenAI's position and that of Anthropic in an interview with POLITICO's Decoded newsletter, whose chief executive Dario Amodei has called for greater caution around the development of the most advanced AI systems.
'I think there's a lot of daylight,' Altman said when describing the difference between the companies. He then argued that the public should accept some negative outcomes if doing so allows people to retain access to the technology.
'We believe that the world should accept some bad things happening for the benefits of this technology and people having the agency,' Altman said.
He later described the types of harm he was unwilling to treat as sufficient reason to impose a restrictive trade-off, including major hacks, misuse and scams. 'I wouldn't take a trade of saying, 'We'll make sure there's no major hacks, there's no misuse of this technology, there's zero scams, there's zero all the other bad things that will happen,'' Altman said.
'Because I think people will do tremendously - orders of magnitude more - good stuff than bad stuff.'
OpenAI AI Regulation Position Is More Complicated
The comments do not amount to a rejection of tighter AI regulation across the board. OpenAI has publicly backed mandatory, capability-based national AI safety regulation and said it expects to support legislation that materially raises safety standards.
In September, OpenAI's chief global affairs officer Chris Lehane said the company wanted to work with Congress on national requirements covering areas including testing, independent assessments, cybersecurity and incident reporting. OpenAI also said it would continue supporting state legislation until Congress acts.
The company has also endorsed several California measures covering independent AI risk assessments, auditor standards, protections for young people using companion chatbots and safeguards against AI-enabled biological threats.
OpenAI's September policy announcement marked a more supportive stance towards state-level AI safety legislation than the company had previously taken.
The difference with Anthropic is therefore more specific than a simple pro-regulation versus anti-regulation split. Anthropic's published policy framework calls for increasingly rigorous external testing, incident reporting and government oversight as AI capabilities and risks rise.
Its proposed rules focus on frontier developers and risks including biological weapons, offensive cyber operations and loss of control over AI systems.
Altman has also objected to the idea of a single laboratory controlling powerful AI and deciding how its benefits should be distributed. His position favours keeping access to the technology broad rather than accepting a model in which control is concentrated in one organisation.
That makes his latest remarks less a call for unrestricted AI than an argument over where regulation should draw its line. OpenAI's own policy documents show the company is seeking government rules and external oversight, while saying that safety should take priority over capability growth when the two cannot be achieved simultaneously.
