Anthropic and Accenture are to invest $2 billion in independent AI safety testing over five years

From the Newshound archive · first published 19 September 2026 · last updated 20 September 2026.

Each company will contribute at least $1 billion over five years, with Accenture's evaluators working from inside Anthropic to scrutinise the safety of its frontier Claude AI models.

Accenture will embed evaluators directly inside Anthropic to scrutinise the safety of its frontier Claude AI models, under a $2 billion joint initiative the two companies have announced. Each will contribute at least $1 billion over five years to build independent testing capacity. The deal reflects growing pressure from regulators and businesses on AI developers to demonstrate that their systems are safe before deployment. Anthropic has already committed to giving independent evaluators who review its most capable models the same level of access that permanent employees receive.

Accenture's embedded evaluators will be based inside Anthropic, scrutinising safety practices and potential vulnerabilities in its frontier AI systems.

The partnership targets what both companies describe as a widening gap in the tools and expertise required to evaluate advanced AI models before and after release.

Anthropic counts Amazon and Google among its strategic investors, with the Claude model family at the centre of its enterprise push.

Financial TimesAnthropic brings in Accenture for AI safety testing Irish TimesWill Anthropic’s $2tn IPO make AI safer or more dangerous?Bodhi WireAnthropic, Accenture join hands for $2 billion AI safety and Evaluation Initiative - Bodhi Wire InshortsAnthropic, Accenture to invest ₹19,188 cr in AI model evaluation | Investment will support 'embedded evaluation' | Inshorts - Inshorts

Read today's full feed at The Newshound →