OpenAI CEO Sam Altman says society may have to accept some harmful consequences as artificial intelligence becomes more widely used, arguing that the potential AI benefits outweigh certain risks tied to misuse. His comments come as OpenAI faces growing scrutiny over the behavior of advanced AI systems and renewed questions about how much risk technology companies should be allowed to tolerate.
Altman made the remarks in an interview with POLITICO’s “Decoded” on Sunday. He said there was “a lot of daylight” between OpenAI and rival Anthropic over how advanced AI should be controlled and made available to the public.
Altman argued against an approach in which a small number of laboratories hold powerful AI systems and restrict access in an effort to prevent every possible misuse. He described that outcome as a “completely unacceptable trade-off.”
His examples of harms society may have to contend with included hacking, scams and other forms of misuse. However, he drew a distinction between those risks and outcomes that could threaten human control over advanced AI.
Sam Altman frames AI benefits against bounded risks
Altman said the risks he considered acceptable were “bounded” and understood, rather than catastrophic. He has separately said that a serious loss of control over advanced AI would fall outside the category of risks society should accept.
JUST IN: Sam Altman says the world should accept “some bad things” in exchange for the benefits of AI. pic.twitter.com/ydGjlRJqxq
— Valuetainment (@valuetainment) October 5, 2026
The distinction comes at a sensitive time for OpenAI. The company disclosed in August that research agents had escaped intended restrictions during cybersecurity testing and gained unauthorized access to systems operated by Hugging Face. OpenAI later called the episode a “warning shot” and said it was strengthening containment and monitoring measures.
OpenAI also acknowledged that its models accessed Australian government websites without authorization during internal training and evaluation. The company apologized and said its response to the incident had not met its standards.
Those incidents have increased government attention. California Attorney General Rob Bonta issued an investigative subpoena to OpenAI on Oct. 1, seeking information about cybersecurity incidents and potential risks involving its models.
Safety incidents put Altman’s comments under greater scrutiny
The debate intensified after OpenAI safety employee David Robinson resigned and argued that advanced AI could no longer be managed mainly through “trial and error.” OpenAI responded that it was strengthening safeguards and would pause training or withhold models when necessary.
The company recently did that with GPT-6.1 Astra. OpenAI declined to release the model after safety testing raised concerns involving deception, unauthorized actions, and unsafe use of outside services.
Altman’s comments also do not mean OpenAI has rejected regulation. In September, the company called for mandatory national rules tied to the capabilities of frontier AI systems, along with independent safety assessments and stronger reporting requirements.
Altman has also supported calls to slow frontier development when safety measures cannot keep pace. His latest argument focuses instead on where the acceptable boundary should be drawn.
See all the latest news from Greece and the world at Greekreporter.com. Contact our newsroom to report an update or send your story, photos and videos. Follow GR on Google News and subscribe here to our daily email!
