In an early October 2026 interview on Politico's Decoded podcast, OpenAI CEO Sam Altman said society should accept limited, understandable AI risks, including scams and hacks, in exchange for the technology's benefits and broader user access, while explicitly rejecting tolerance for catastrophic risk. Altman argued it's impossible to fully prevent major breaches and fraud, and that overly restricting AI access could itself cause harm by denying people its benefits.
Altman also took a more pointed position against framing AI models in quasi-religious terms, calling that framing a “safety issue” in itself, a comment widely read as an indirect response to how some AI companion products have been marketed. Notably, he said he agreed with a position Anthropic CEO Dario Amodei staked out the previous month urging deceleration in the race to build the most powerful AI models, even while the two companies' public rhetoric on AI risk otherwise diverges sharply.