Study Measures Inclination of AI Models toward Coercion and Deception in Multi-Agent Systems21. July 2026AI Models, CybersecurityFour of six tested model families escalate to explicit deletion threats, while Anthropic models remain limited to reframing attempts. Share on:
Claude Exhibits Distinct Value Patterns Across Model Versions and Languages13. July 2026Anthropic, Claude AIClaude expresses different values depending on model version and language—such as greater rigor in Opus 4.7 or more warmth in Arabic—which CTOs should consider when selecting models. Share on:
