Anthropic and Accenture said they would each commit at least $1 billion over five years to build capacity for independent evaluation of frontier AI models.
The partnership will be led by Faculty, Accenture’s specialist AI business, and is expected to include model evaluations, red teaming, alignment assessments and testing of safeguards. Anthropic calls the approach “embedded evaluation”: outside evaluators would work inside AI companies with access comparable to employees.
Anthropic says the model is new and that many details of how it will operate are still being worked out. That leaves an important question unresolved: how independent an evaluator can remain while working inside, and being funded by, the company it is evaluating.
The partnership is significant less because it settles that question than because it gives outside evaluation money and institutional scale. Its credibility will depend on the access evaluators receive, the freedom they have to report adverse findings and the transparency of the resulting assessments.
AI Safety Watch 