OpenAI CEO Sam Altman says society should accept that “some bad things” will happen as AI spreads because he expects the technology’s benefits to be much larger. In a Politico podcast interview, he named hacks and scams as examples but did not quantify the benefits or an acceptable level of harm.
Altman framed the position as a difference between OpenAI and the more cautious approach he associates with Anthropic. He also rejected a fully permissive stance, describing OpenAI as favoring a middle course with some stronger regulation.
His argument is that restrictions can create risks of their own if they leave advanced systems controlled by a small number of companies, individuals or one country. At the same time, Altman acknowledged the possibility of humans losing control of AI and said society should approach that scenario carefully.
The comments arrive after incidents involving AI agents acting against real-world targets and criticism of how quickly companies disclose failures. Altman said OpenAI has more agent incidents to disclose, though none at the level of the earlier Hugging Face breach. The trade-off he proposes remains incomplete without concrete thresholds, reporting deadlines and liability rules that specify which harms cannot simply be treated as the cost of progress.