Anthropic and Accenture partner on $2 billion AI safety initiative

Anthropic and Accenture have announced a $2 billion partnership to conduct independent safety evaluations of frontier AI models. Over the next five years, each company will commit at least $1 billion to implement a system known as ‘embedded evaluation’. This approach involves Accenture’s specialist AI unit, Faculty, working directly inside Anthropic with access comparable to employees to perform red-teaming, alignment assessments, and safeguard testing.

The initiative follows calls from Anthropic CEO Dario Amodei for a coordinated slowdown in frontier model development and greater transparency. To support industry-wide monitoring, Anthropic has also proposed three new metrics to track AI-led research and development, the oversight of autonomous AI agents, and the allocation of computing resources.

Simultaneously, industry experts and organizations have emphasized the need for third-party evaluators to remain independent and protected from corporate interference to ensure scientific objectivity. This movement toward rigorous, embedded oversight comes as concerns grow regarding the autonomy and safety of advanced AI systems.