Anthropic says it met Responsible Scaling Policy for recursive self-improvement

In a series of S-1 filings leading up to its public market debut, AI lab Anthropic revealed that it has met its Responsible Scaling Policy threshold for recursive self-improvement. The company stated that its internal estimates and expectations for future models indicate a dramatic acceleration in AI research and development capabilities driven by AI systems building AI. Following this milestone, Anthropic has promised to increase investments in safety R&D, systematic alignment assessments, and internal red teaming. Open article →