AI Trains Safer AI, Outperforming Human Safety Researchers

Original: Automated Researchers Mitigate Alignment Failures

Source: anthropic.com ↗