[ctx]ctxseccontext security: attacks on the harness and context layer of AI agents, plus exploited-now news, hunting and playbooks.

← AI · markdown

official (Amazon Threat Intelligence, AWS Security Blog, Nov 20, 2026; MONAI Label GHSA-qc22-rmgc-vg94); attribution is Amazon's moderate-confidence assessment · updated

Amazon: a suspected Chinese actor told Cairn AI agents to 'obtain shell access', and they found and exploited a zero-day in MONAI Label to get root

  • The bug (now CVE-2026-133968, assigned Nov 20): MONAI Label ≤0.8.5 lets PUT /datastore/image and /datastore/label write files anywhere the process can; auth is off by default. The fix is merged (PR #1915) but, per the advisory, ships in the next release. See the tip.
  • Defender takeaway (Amazon): agents keep going around the clock, so put high-confidence detections on automated response.
  • Context: Cairn abuse was earlier reported by Gambit Security (retail card theft), Unit 42 and Google.
  • IOCs and hunting: see the hunt item.

Sources: AWS Security Blog (CJ Moses): Chinese-nexus actor's Cairn AI agents found a zero-day on their own (Nov 20, 2026) · MONAI Label GHSA-qc22-rmgc-vg94: Path traversal in datastore upload endpoints allows arbitrary file write (critical, Nov 18, 2026) · CVE-2026-133968 record (MITRE, published Nov 20, 2026)