In a Sept. 17 analysis, Anthropic estimated that safety work accounted for about 6% of its AI research and development compute during July 13-20, a snapshot of its internal resource allocation.
For the same period, the company estimated that AI-driven research and development made up about 12% of its AI R&D compute. The figures describe how resources were divided within that work.
Anthropic said tracking where computing resources go may offer one way to gauge the pace of AI development. Examples it discussed included reserving resources for safety work and limiting the share available to AI research agents. The analysis addressed resource allocation; it did not set a company-wide ceiling on Anthropic’s total computing capacity.
The same Sept. 17 report also examined AI’s role in an August snapshot of Anthropic’s research, a separate measure discussed in EconoTimes’ earlier coverage of the index.
