Anthropic Picks Accenture for $1 Billion AI Oversight Push as It Tests Slowdown Plan
Updated
Updated · CNBC · Sep 18
Anthropic Picks Accenture for $1 Billion AI Oversight Push as It Tests Slowdown Plan
3 articles · Updated · CNBC · Sep 18
Summary
Accenture will become Anthropic’s first embedded evaluator, with Faculty staff getting employee-level access to test safeguards, red-team models and assess whether systems align with human values.
At least $1 billion from Anthropic and Accenture is earmarked over five years to build that evaluation capacity, though Anthropic said it will directly fund the work for now because no pooled or government funding exists yet.
The move is the first concrete implementation of CEO Dario Amodei’s three-step proposal, unveiled last Saturday, to slow the pace of advanced AI development amid warnings that frontier models could cause catastrophic harm.
Anthropic said the arrangement is nonexclusive and it is also in talks with nonprofit METR and other third parties, while stressing that outside evaluators will not reduce the company’s own responsibility for model safety.
The announcement gives practical shape to a plan that drew support from OpenAI’s Sam Altman and Elon Musk but skepticism from Nvidia’s Jensen Huang, as Anthropic faces scrutiny while preparing for an expected IPO.