Anthropic’s Daniela Amodei believes the market will reward safe AI

Trump administration It may be thought that regulation is Disabled the Artificial intelligence industryBut one of the biggest players in the industry disagrees.

At WIRED’s big interview event on Thursday, Anthropic Although Trump’s A.I. and Cryptocurrency czar David Sachs, may be chirp Since her company “runs a sophisticated regulatory oversight strategy based on fear-mongering,” she is convinced that her company’s commitment to disclosing the potential risks of AI makes the industry stronger.

“We were very vocal from day one that we felt there was this amazing potential” for artificial intelligence, Amodei said. “We really want to be able to make the whole world realize the potential and the positive benefits and the upside that can come from AI, and in order to do that, we have to get the hard things right. We have to make the risks manageable. That’s why we talk about it a lot.”

More than 300,000 startups, developers and companies use some version of Anthropolic’s Claude model, and Amodei said she has learned, through the company’s dealings with those brands, that while customers want their AI to be able to do great things, they also want it to be reliable and secure.

“Nobody is saying, ‘We want a less safe product,’” Amodei said, likening Anthropic’s report on the limitations of its model and jailbreak to that of an auto company launching crash-test studies to show how it addresses safety concerns. It may seem shocking to see a crash test dummy fly through a car window in a video, but knowing that an automaker updated their car’s safety features as a result of that test might sell a buyer on the car. The same goes for companies using Anthropic’s AI products, creating a somewhat self-regulating market, Amodei said.

“We are setting what would be considered minimum safety standards just by what we are putting into the economy,” she said. “Companies are now building a lot of workflows and everyday tool tasks around AI, and they say, ‘Okay, we know this product doesn’t hallucinate as much, doesn’t produce harmful content, doesn’t do all these bad things.’ Why would you pick a competitor that’s going to get a lower score on that?”

Photo: Annie Noelker