Governance, Identity, Security, and Incident Disclosure Become the Adoption Gate

Enterprise agent risk is expanding from model safeguards to identity, authorization, monitoring, MCP and supply-chain trust, policy enforcement, rollback, and independent oversight. GitHub's AI Controls, Okta's practical security frame, and hands-on reports about unsafe local-agent access reinforce that governance requires isolation and continuous telemetry, not approval prompts alone. AIR's $50M financing and emerging runtime-security products signal a distinct governance category.