Accenture, Anthropic commit $1B each to AI safety testing

Accenture ↘ 1.76% and Anthropic announced on Friday a partnership to build a team of embedded evaluators who will work inside Anthropic to assess the safety of its frontier AI models, with each company committing to invest at least $1 billion over five years in AI safety initiatives.investing+1

The deal marks one of the largest corporate commitments to AI safety evaluation and introduces a novel structure in which outside evaluators receive employee-level access to an AI developer's internal operations.

How Embedded Evaluation Works

Unlike traditional external auditors, embedded evaluators will sit inside Anthropic's offices with access badges, company laptops, and permissions comparable to those of internal risk-assessment teams, according to the announcement. That access allows evaluators to observe models during training, follow decisions governing how models are built and deployed, and speak directly with employees.unite

The work will be led by Faculty, a specialist AI firm Accenture acquired that has experience testing models for leading AI labs and building systems for government and defense clients, including the UK National Health Service's Early Warning System during the COVID-19 pandemic.finance.yahoo

"Safety requires both deep technical expertise and a clear understanding of how AI is used in the real world," said Julie Sweet, chair and CEO of Accenture. "Embedded evaluation is an emerging area and we look forward to partnering with Anthropic to help accelerate the development of embedded evaluators."finance.yahoo

Transparency and Publication Rights

Under terms described in a September 2026 essay by Anthropic CEO Dario Amodei titled "We Must Pace the Frontier," external reviewers will hold the right to publish findings about risk levels, incidents, and practices without editorial control by Anthropic. The company retains a narrow ability to redact security-sensitive or legally privileged material but cannot suppress unfavorable conclusions. Reviewers may publicly disclose if a redaction removed something material to their findings.unite

In the essay, Amodei outlined a three-step plan — embedded evaluators, coordination among frontier AI companies in democratic countries, and global coordination — and called on governments to require other frontier labs to adopt similar measures. "We must slow the pace at which we improve the capabilities of AI models," he wrote.unite

A Non-Exclusive Arrangement

The partnership is non-exclusive. Anthropic said it plans to announce additional evaluators in the coming weeks and is in dialogue with METR and other nonprofit organizations to pilot elements of embedded evaluation. The arrangement extends a broader relationship formed in December 2025, when the two companies launched the Accenture Anthropic Business Group and began training roughly 30,000 Accenture professionals on Anthropic's Claude models.unite

Accenture's stock rose in early trading following the announcement.seekingalpha