Anthropic just outsourced part of its safety homework to a consulting firm best known for helping enterprises roll out software, not break it.
The company has picked Accenture to run third-party evaluations of its AI systems' safety. Anthropic is framing the deal as more than a vendor contract: it's described as the first concrete step in a plan, driven by the CEO, to deliberately slow the pace of AI development at the company. No further details on scope, timeline, or what "slowing down" will actually look like in practice have been specified.
The pairing is the interesting part. Frontier AI safety testing has typically lived with in-house red teams or specialized outside evaluators, not Big Four consultancies whose core business is implementation and change management for corporate clients. Bringing in a firm like Accenture signals Anthropic wants safety review to look like a standard, auditable business process rather than a research project done behind closed doors. Whether that buys more rigor or just more paperwork is the open question.
A safety evaluation contract is an easy thing to announce and a hard thing to verify from the outside. "Slowing down AI development" is a bold framing for hiring an outside auditor - the real test is whether it changes what ships, and when.