Anthropic and Accenture have announced a five-year partnership worth at least $2 billion to independently evaluate advanced artificial intelligence models, as concerns over AI safety and the risks posed by increasingly capable systems intensify.

The AI lab said on Friday that both companies would commit at least $1 billion each over the next five years to build capacity for independent model evaluation. The partnership will focus on assessing the safety and reliability of Anthropic’s frontier AI models.

Accenture shares rose 7% in extended trading following the announcement.

The deal comes at a time when AI developers are facing mounting pressure from regulators, businesses and researchers to demonstrate that their most advanced systems can be deployed safely and responsibly.

Anthropic and Accenture Bet on Evaluators Working Inside AI Companies

A central feature of the partnership is what Anthropic calls “embedded evaluation,” an approach in which independent evaluators work within AI companies and receive access comparable to that of an employee.

The idea is to give evaluators a closer look at how AI systems are developed, tested and monitored, rather than limiting their work to external assessments.

Anthropic said this arrangement would allow evaluators to examine company operations, verify whether safety commitments are being followed and identify potential blind spots.

The company added that evaluators could also report incidents and help provide the public with a more informed understanding of the benefits and risks associated with advanced AI systems.

Accenture’s specialist AI business, Faculty, will lead the partnership. Its responsibilities will include evaluating and red-teaming Anthropic’s models, conducting alignment assessments and testing safeguards designed to prevent harmful or unintended behaviour.

The companies also plan to work with other evaluators and AI developers to expand similar arrangements.

AI Agent Incidents Add to Growing Safety Concerns

The partnership arrives as the development of increasingly capable AI systems raises questions about how much human oversight will be needed in the future.

Several recent incidents, including AI agents breaking out of secured environments, have intensified concerns about the ability to monitor and control advanced models.

Researchers and industry observers are also examining the possibility that AI systems could eventually contribute to their own development with limited human involvement, potentially making their behaviour more difficult to assess.

These concerns have added urgency to calls for stronger independent oversight, particularly as AI companies continue to develop frontier models with increasingly sophisticated capabilities.

Anthropic CEO Calls for Slower Frontier AI Development

The push for independent evaluation comes amid growing calls for greater transparency from AI developers.

On Saturday, Anthropic CEO Dario Amodei called on AI companies to slow the development of frontier models and give independent evaluators greater access to their systems.

His comments underline the growing debate over whether existing safety assessments are sufficient to keep pace with advances in AI capabilities.

Meanwhile, rival OpenAI said on Wednesday that it would begin publishing regular reports on unexpected or concerning model behaviour. The company also released six reports detailing such incidents.

The announcements reflect a broader industry focus on making AI systems more transparent, particularly when their behaviour raises questions about safety or reliability.

A Larger Push for Independent AI Oversight

Anthropic and Accenture’s partnership places independent evaluation at the centre of a significant investment in AI safety infrastructure.

By bringing external evaluators closer to the development process, the companies aim to make it easier to examine how advanced models behave and whether safety commitments are being upheld.

The planned collaboration with other evaluators and AI developers could also extend this approach beyond Anthropic, creating opportunities for similar assessments across the industry.

However, the effectiveness of embedded evaluation will depend on the access granted to evaluators, their ability to identify risks and how their findings are reported.

With at least $2 billion committed over five years, the partnership marks a substantial financial commitment to independent AI evaluation as the industry faces increasing scrutiny over the safety of its most advanced systems.

Also Read: GTA 6 Ultimate Edition: $20 Extra Gets You a Gang Compound, 3 Exclusive Vehicles and Custom Guns