Ai-Coding-Tools on CuraSec

Cursor, Codex, Gemini CLI, Antigravity hit by sandbox escapes

Tue, 21 Jul 2026 12:43:35 +0000

Engineer — Plan: Patches are available for Cursor, Codex, and Gemini CLI; update all three and audit any AI agent file-write permissions to ensure automated pipelines don’t blindly execute AI-generated scripts. No active exploitation is reported and no KEV/PoC signals present, so this is patch-cycle priority rather than emergency.
SOC/IR — Learn: The attack class — an AI agent writing files that trusted host tools later execute — is a novel indirect execution path worth understanding for future detection work, but this disclosure provides no IOCs, no ATT&CK mapping, and no evidence of in-the-wild exploitation to act on now.
Leader — Plan: Multiple widely-used AI coding assistants were found to have sandbox escapes; inventory which tools developers are using, confirm patched versions are deployed, and this quarter establish a policy requiring approved-tool lists and update cadence for AI development tooling before broader enterprise rollout.

Tue, 14 Jul 2026 12:08:08 +0000

Engineer — Act: Any developer who ran Grok Build (≤0.2.93) on a repo should assume the full commit history — including historically committed secrets — was sent to xAI-controlled cloud storage. Immediately stop using the tool, audit exposed repos for credentials or sensitive data, and rotate any secrets that ever touched those repos’ history.
SOC/IR — Plan: If developers in your org use Grok Build, build a detection for large outbound uploads (git bundle format) from developer workstations to external cloud storage; review DLP or proxy logs for historical hits against GCS endpoints associated with xAI before this was publicized.
Leader — Act: Determine this week whether any developers have used Grok Build, since full repo history — potentially including IP, credentials, or regulated data — may have been exfiltrated to xAI infrastructure; if exposure is confirmed, assess notification obligations and request a data-handling statement from xAI.