Anthropic reports that artificial intelligence now leads 26% of its research and development operations. This marks a significant shift from the beginning of the year, when such practices were not yet implemented. Furthermore, AI is currently involved in 90% of collaborative tasks alongside human researchers.
As of August 2026, more than 30,000 AI agents are simultaneously engaged in engineering and research work within the company. Anthropic allocates between 6% and 12% of its total computing resources specifically for security monitoring tasks to ensure safe development processes.
The company advocates for recursive self-improvement of AI systems while emphasizing the need for transparency. Anthropic has expressed readiness to involve third-party auditors to evaluate the safety of its development processes and has proposed standardized criteria for monitoring AI agent activity.
Anthropic maintains that no AI agent currently operates with full autonomy in research and development. The company urges the industry to adopt monitoring systems that allow for human intervention and to correlate resource allocation data with practical performance testing results.
