Frontier AI Risk Governance

As frontier AI models approach capabilities like recursive self-improvement, international governance efforts define 'red lines' and catastrophic risk thresholds to prevent uncontrolled escalation. Together these terms capture the technical risk (models improving themselves beyond oversight) and the policy response (agreed limits before that risk materializes).