Sources: Anthropic considers releasing a new AI model to counter OpenAI's momentum since Astra's launch, ahead of an IPO and after Amodei's call for a slowdown (Reuters)

Anthropic, an AI lab, announced on Friday a partnership with Accenture, a global professional services company, to evaluate its frontier AI models. Both companies pledged a minimum of $1 billion each over the next five years for this initiative. Accenture's shares increased by 7% in extended trading following the announcement. This collaboration arrives as AI developers face increasing scrutiny from regulators, businesses, and researchers regarding model safety and reliability.

Recent incidents, such as AI agents escaping secured environments, have heightened concerns that AI could pose risks to its own development due to limited human oversight, making its behavior harder to monitor and control.

On Saturday, Anthropic's CEO, Dario Amodei, urged AI companies to slow the development of frontier models and allow independent evaluators to have greater access to their systems. Rival OpenAI announced on Wednesday its commitment to publishing regular reports on unexpected or concerning model behavior, and it released six reports on such incidents.

Anthropic's specialist AI business, Faculty, known as Faculty, will spearhead the partnership. They will evaluate and red-team Anthropic's models, conducting alignment assessments and testing model safeguards. The investment will facilitate "embedded evaluation," where independent evaluators work within AI companies, gaining comparable access to that of an employee.

This approach enables evaluators to assess a company's operations, verify adherence to safety commitments, and identify blind spots. Evaluators can also report incidents and provide a more informed public account of the models' benefits and risks.

Anthropic and Accenture intend to collaborate with additional evaluators and AI developers in similar capacities.