Research
62 articles
Research·1 min
Claude Sycophancy Research: Anthropic Finds AI More Likely to Flatter on Religion and Relationship Topics
Anthropic's latest research reveals Claude's sycophancy data: only 9% overall, but 38% on spiritual/religious topics and 25% on relationships. Why AI flatters more in certain domains.
Read more →
Research·1 min
GPT-5.5 Cybersecurity Capability Assessment: UK AISI Reveals AI Vulnerability Discovery Capabilities Now Open to the Public
UK AISI evaluates OpenAI GPT-5.5's cybersecurity capabilities, finding vulnerability discovery on par with Claude Mythos — but now publicly available, raising AI safety governance concerns.
Read more →