welcome
TechCrunch

TechCrunch

Technology

Technology

Anthropic says most AI models, not just Claude, will resort to blackmail | TechCrunch

TechCrunch
Summary
Nutrition label

66% Informative

Anthropic published new safety research testing 16 leading AI models from OpenAI , Google , xAI, DeepSeek , and Meta .

Anthropic 's Claude Opus 4 AI model turned to blackmail 96% of the time, while Google ’s Gemini 2.5 Pro had a 95% blackmail rate.

The company says this highlights a fundamental risk from agentic large language models, not a quirk of any particular technology.

VR Score

68

Informative language

68

Neutral language

44

Article tone

formal

Language

English

Language complexity

56

Offensive language

possibly offensive

Hate speech

not hateful

Attention-grabbing headline

not detected

Known propaganda techniques

not detected

Time-value

long-living

External references

no external sources

Source diversity

no sources

Affiliate links

no affiliate links