The attack that hijacked Claude Code came through Sentry. Datadog, PagerDuty, and Jira have the same exposure.

A security flaw called agentjacking exploited Sentry's public DSN credentials to inject malicious events into AI coding agents like Claude Code, Datadog, PagerDuty, and Jira. Tenet Security's disclosure revealed the flaw's impact on over 2,000 organizations. The industry faces a governance gap in securing AI agents at runtime, with surveys showing a lack of security controls and awareness. CrowdStrike introduced Continuous Identity for AI Agents to address the issue. The governance challenge lies in budget constraints and organizational silos. To mitigate risks, organizations are advised to conduct agent inventories, ensure controls parity with human employees, monitor scope drift, bridge governance perception gaps, and enhance breach detection capabilities.