North America

Autonomous AI needs safeguards beyond model-level guardrails, study finds

Emergence’s study of autonomous agents found that systems powered by different large language models can produce sharply different long-run behaviours under identical rules.