Anthropic and Accenture invest $2 billion to enhance AI safety assessments

On September 20, 2026, Anthropic, an artificial intelligence laboratory, announced a partnership with Accenture to conduct independent safety assessments of its advanced AI models. The collaboration will see both companies invest at least $10 billion over the next five years, equivalent to approximately 67.18 billion yuan. This initiative aims to build the necessary infrastructure for ongoing safety evaluations. Anthropic CEO Dario Amodei called for a slowdown in the development of cutting-edge models and greater access for independent assessors. Meanwhile, OpenAI has already begun publishing reports on model anomalies and released six incident reports. Accenture’s AI division, Faculty, will lead the assessments, including red-blue team exercises and safety mechanism testing. The partnership will support an embedded evaluation model, where assessors work within Anthropic’s facilities with access levels similar to regular employees. This approach allows for a deeper inspection of operational practices and potential safety blind spots. Assessors will also report incidents and provide detailed public analyses of AI risks and benefits. Anthropic plans to expand this model with other assessment bodies and AI developers.