Anthropic published a paper titled “Automated Researchers Can Reliably Mitigate Alignment Failures,” led by fellow Chen Yueh-Han. The system replicates the traditional research process: each automated system searches literature, proposes a method, and trains the model with that method for 30 minutes, iteratively improving benchmark scores. Effective methods are kept; ineffective ones are discarded. Given 10 benchmarks for specific misaligned behaviors, the system improved performance on every one without degrading overall performance.

The paper positions this as early evidence that automated alignment post-training could become practical in the near term. The system is explicitly compared to its human equivalent, the Automated Alignment Researcher (AAR). “The best AAR method beats what experienced humans propose, on average within six hours,” the paper states. “Human guided research directions do not lead to stronger performance.” A cost comparison is also provided: “An AAR costs roughly $4 per hour in API inference against the $150 per hour we pay our human researchers.”

The paper acknowledges limitations: the system only works if benchmarks reflect actual alignment goals, and significant work remains in establishing and maintaining those benchmarks, as well as sustaining the literature the automated researchers draw from.