Anthropic co-founder and CEO Dario Amodei speaks at the Inbound 2025 Powered by HubSpot artificial intelligence panel at Moscone Center in San Francisco on September 4, 2025.
Chance Ye | Getty Images Entertainment | Getty Images
Anthropic is Friday. Accenture The company’s work as an embedded evaluator is the company’s first concrete step toward implementing CEO Dario Amodei’s proposal to slow the pace of artificial intelligence development.
Anthropic and Accenture have agreed to invest at least $1 billion over the next five years to “build capacity in this area,” according to the release. But Anthropic said it would provide direct funding to Accenture’s work given its “materiality and urgency.”
“Longer term, we believe that funding should come from pooled or government sources, as we called for in the Advanced AI Framework in June,” Anthropic said. “We will be working with different valuers under different funding arrangements as neither currently exists.”
Amodei rocked the tech world on Saturday by announcing a three-step plan to fine-tune how quickly AI companies improve their cutting-edge models. Anthropic and its main competitors, OpenAIAI has come under intense scrutiny in recent weeks as researchers increasingly warn of its potential to cause catastrophic harm.
Amodei’s proposal was announced this week by OpenAI CEO Sam Altman, tesla and space x Not just CEO Elon Musk, but others as well. Nvidia CEO Jensen Huang dismissed the concerns, insisting no new regulations were needed. Many questioned what Amodei’s proposal actually meant, especially as Anthropic prepares for what is widely expected to be a major IPO.
The first step in Amodei’s plan is to give third-party evaluators employee-level access to the company to verify safety practices and report incidents. He said Saturday that Anthropic had “unilaterally” committed to this part of the proposal and encouraged other AI companies to do the same.
Anthropic announced Friday that it has initially agreed to bring in employees from Accenture’s AI specialist business, Faculty, to test safeguard and red team models and assess whether the models behave in line with human values. The partnership is not exclusive, and Anthropic said it is in discussions with research nonprofit METR and other third parties.
The company emphasized that it is still responsible for the safety of its models and said that working with built-in evaluators does not alleviate that responsibility.
“We are now sharing these early efforts so that people and other AI developers can see our process,” Anthropic said. “We expect our approach to evolve as the field matures, and we will share more once our work begins and we welcome additional evaluators.”
See: Creating an AI Safety Handbook
