Anthropic Picks Accenture as First Embedded AI Evaluator in $1B Safety Push

Anthropic has selected Accenture as its first embedded evaluator to help slow the pace of artificial intelligence development, advancing a three-step proposal outlined by CEO Dario Amodei on September 12. The proposal calls for independent evaluators with employee-like access to AI labs, giving them the ability to assess models, conduct red-teaming exercises, and test safeguards before potentially catastrophic systems emerge. OpenAI CEO Sam Altman and SpaceX CEO Elon Musk responded positively to the plan, though Nvidia CEO Jensen Huang pushed back, arguing that additional regulation was unnecessary.

The partnership will focus on evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards, leveraging Accenture's AI division, Faculty, which specializes in testing models for leading AI labs. Both companies have committed at least $1 billion over the next five years to the initiative, with Anthropic funding the work directly given the urgency. Amodei has acknowledged that no existing system covers long-term funding for independent evaluation and suggests pooled or government sources should eventually shoulder the cost.

"Embedded evaluation is an emerging area, and we look forward to partnering with Anthropic to help accelerate the development of embedded evaluators, which we see as an important part of the safety landscape going forward," said Accenture Chair and CEO Julie Sweet. The collaboration is non-exclusive, and Anthropic confirmed it expects to announce additional evaluator partnerships in the coming weeks. Amodei has warned that AI's capacity for recursive self-improvement could outpace humanity's ability to understand and control these systems if development continues unchecked.