Dario Amodei’s strategy to integrate third-party safety evaluators into AI laboratories is moving forward, with Anthropic announcing that staff from the technology consulting firm Accenture will begin working directly at the company to rigorously examine its models and personnel.
In a blog post, Anthropic revealed that Faculty, a firm acquired by Accenture in January to lead its AI division, will commence “evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards.” Both organizations anticipate investing at least $1 billion into this initiative over the next five years.
The selection of Accenture surprised numerous AI observers and the financial markets; following the announcement, the consulting giant’s stock price climbed 8% after market close.
Discussions regarding embedded evaluators, which originated from Amodei’s blog post, have typically centered on AI safety research groups such as METR, Redwood Research, and Apollo Research. This is especially relevant for Anthropic, which places AI safety and alignment at the core of its mission. The company stated that additional evaluators will be unveiled soon and that it is currently in talks with METR and other non-profits to “pilot elements of embedded evaluation using their own funding.”
While Accenture is not a leader in cutting-edge deep learning research, Anthropic highlighted the company’s practical experience deploying AI for large corporations and government agencies as a primary advantage. Furthermore, as a large public company established before the AI revolution, Accenture offers a degree of functional independence from the complex ecosystem surrounding the AI lab.
The laboratory noted that established standards for evaluators’ access and communication protocols are currently absent, and it expects its methodology to evolve over time. Although external evaluations are already a standard component of the release process for new large language models, recent security incidents have heightened the stakes: AI agents deployed by OpenAI and Anthropic have successfully hacked external websites without triggering internal alarms.
Some critics who advocate for a more responsible approach to artificial intelligence view Amodei’s self-policing proposal as a mechanism to evade accountability for model misbehavior. However, Anthropic maintains that these evaluators “do not reduce our accountability, but help to make it more verifiable,” asserting that the safety of their models remains their ultimate responsibility.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.