Anthropic picks Accenture as embeded evaluator

Anthropic and Accenture will invest $2B over five years to evaluate AI models. Faculty, Accenture's AI division, will lead the project, focusing on model safety and security. Anthropic's CEO Dario Amodei previously advocated for slowing AI development and independent checks. The partnership is non-exclusive, with plans for more collaborations. Anthropic acknowledges potential conflicts in funding evaluations.

Accenture and Anthropic have partnered to create a team of embedded evaluators to test and assess Anthropic's AI models, with both companies committing to invest at least $1 billion each in AI safety over the next five years. The team will evaluate models, conduct alignment assessments, and test safeguards. Accenture will leverage its acquisition of Faculty, an applied AI company with experience in safety and ethics.

Anthropic selected Accenture as its first embedded evaluator to slow AI development, per CEO Dario Amodei's proposal. Both companies plan to invest at least $1 billion over five years. Accenture's AI business, Faculty, will assess and test Anthropic's models. Funding will initially come from Anthropic, with long-term sources yet undetermined.

Anthropic and Accenture partner to create an independent AI safety evaluation framework, investing at least $1B each over five years. The collaboration will focus on evaluating AI models, with embedded evaluators providing deeper scrutiny. Anthropic seeks to develop industry standards and is exploring additional partnerships.

Anthropic has chosen Accenture as its first embedded evaluator, with both companies committing at least $1bn over five years. Anthropic will fund Accenture directly but advocates for pooled or government funding. Accenture will assess Anthropic's models, with access comparable to Anthropic's staff. Anthropic is also in talks with nonprofit evaluators like METR. The lack of standards for evaluator access and reporting remains an open question.

Accenture and Anthropic plan to invest $1 billion each over five years in AI safety. Accenture will leverage its AI expertise and its subsidiary Faculty's experience in testing AI models. The partnership aims to accelerate the development of embedded evaluators for AI safety, according to company executives.