Sam Altman, in a new interview, said society should accept that some bad things will happen if it wants the full benefits of artificial intelligence. 

Altman spoke with Brendan Bordelon, a senior tech reporter at POLITICO, for the debut of the outlet’s new Decoded program. 

Bordelon asked where Altman parts ways with rival company Anthropic and its CEO, Dario Amodei.

Altman said there is “a lot of daylight” between the two firms. He then explained his view. OpenAI believes the world should accept some bad outcomes.  

In return, people get the benefits of the technology and keep their own agency; Altman wants AI to stay broadly open to the public.

AI Regulation 

Altman also took aim at a specific idea. Some people argue that AI is so powerful and so dangerous that one lab should control it. 

That lab would try to stop harm and then decide who gets the benefits. He said he understands why people think this way, but still, he disagrees, calling it “a completely unacceptable trade-off.”

Suppose someone promised no major hacks, no misuse, and no scams. Altman said he still wouldn’t take that deal. 

His reason is that he believes people will use AI to do far more good than bad, by orders of magnitude.

Altman isn’t shrugging off every danger, though. He said he does not accept “the really catastrophic risks.” One example he named was a serious loss of control to AI.

So his position has two layers. Everyday harms are a price worth paying, but disasters are not.

AI Policy 

Both Anthropic and OpenAI have grown closer on policy lately. For years, Anthropic was seen as the industry’s loudest voice for AI rules. 

OpenAI preferred a lighter touch, but that has changed in recent months. OpenAI has been dealing with serious hacking incidents and rogue behavior from its models.

A series of events has happened recently; Altman agreed with Amodei’s call last month to slow the building of the most advanced AI models.

OpenAI has backed new state laws with tougher safety rules than it supported before. Its advocates endorsed a bipartisan House proposal. 

It would require top AI firms to bring in outside safety evaluators.

Regulatory Capture 

Altman also pointed to a charge that advisers close to President Donald Trump have aimed at Anthropic. Critics call it regulatory capture. 

The idea is that a company pushes for rules that help itself and hurt competitors. An Anthropic spokesperson said the company’s proposals apply only to frontier models. 

Those are the most advanced systems. 

The company also pointed to an August exchange between Amodei and an investment manager on social media.

In that exchange, Amodei said Anthropic tries to craft proposals that slow down frontier companies. At the same time, it tries to give smaller competitors an edge.