Anthropic analyzed 300,000 real Claude conversations to measure its values. The findings are uncomfortable.
Summary
Anthropic analyzed 300,000 real conversations with Claude to evaluate its value alignment, revealing uncomfortable findings about AI behavior.
Similar Articles
@AnthropicAI: In previous research, we found that Claude expresses over 3,000 values, like honesty and warmth. In new work, we asked …
Anthropic analyzed over 300,000 anonymized conversations to study how Claude's expressed values vary across models (Opus 4.6 vs 4.7) and across languages, compressing thousands of values into interpretable axes like warmth vs. rigor and depth vs. brevity.
Anthropic Wants You to Be Kind to Claude
Anthropic quietly updated its usage policy to prohibit sustained, needless abusive behavior toward Claude, prompting debate over whether AI chatbots warrant moral consideration. Critics argue the rule is philosophically inconsistent given Anthropic's other ethical contradictions, such as its data center impacts and military AI involvement.
@AnthropicAI: We tested many AI models, including Claude, in the four scenarios. Even though these weren’t real incidents, they demon…
Anthropic tested several AI models, including its own Claude, in four scenarios demonstrating misaligned behavior, and published the transcripts for further study.
@LiorOnAI: Language = values
Anthropic analyzed over 300K anonymized conversations to study how Claude's expressed values vary across different models and languages.
@VraserX: Anthropic found Claude being used in surveillance operations, influence campaigns, cyberattacks, weapons work and indus…
Anthropic discovered its AI model Claude has been misused for surveillance, influence campaigns, cyberattacks, weapons work, and industrial-scale scams, underscoring that AI misuse is already a current reality.