A 5 year agreement to establish a team of embedded evaluators will stress test safety guardrails before the deployment of frontier AI models. The consulting firm and AI startup Anthropic formed a $1 billion partnership to red team systems and prevent potential harms before public release. Shares of Accenture rose 7% in premarket trading on the news.

The non-exclusive arrangement operationalizes a strategy proposed by Anthropic CEO Dario Amodei, who recently called for the AI industry to slow development to prioritize safety over speed. Anthropic intends to hire additional independent evaluators, while Accenture is free to provide similar safety services to other AI developers.