Do AI agents become riskier under pressure? by Agentic Workforce Key Takeaways Propensity (not capability) determines real AI risk . Models that score well on intelligence tests can still behave unsafely under pressure. Operational pressure dramatically erodes safe... AI Ethics AI Safety AI Security Cyber Security Responsible AI Risk Management Dec 23, 2025