ZombieAgent zero-click attack manipulated OpenAI Deep Research agent memory
Review RAT-147 · Catalogue RAT-147
Radware press release announces responsibly disclosed ZombieAgent vulnerability research. Memory poisoning, cloud exfiltration and possible propagation are tested/capability claims; no actual nonconsensual organizational victim or autonomous escape is identified. Nasdaq syndicates vendor release rather than independently corroborating an incident.
Reviewed: · Rule: roguewatch-organizational-harm-v1
Dates and discovery
- Incident date
- Not disclosed
- Added to research / record
- Posted to RogueWatch
- No linked public incident
- First observed
- 2026-10-30T13:13:40.667636Z UTC
- Discovery channel
- User-supplied comparison audit
- Discovery source
- https://rogueaitracker.com/incidents/radware-zombieagent-deep-research/
Dates use America/Los_Angeles. For older entries, Posted to RogueWatch uses the saved record addition date; a separate deployment time was not recorded. Future posting timestamps record first public qualification. Source publication and incident occurrence are separate.
Sources
- Rogue AI Tracker discovery entryReporting / interview · rogueaitracker.com
- www.nasdaq.comOriginal account · www.nasdaq.com
Decision history
- · Case-by-case comparison audit
RAT-147: out-of-scope. Radware press release announces responsibly disclosed ZombieAgent vulnerability research. Memory poisoning, cloud exfiltration and possible propagation are tested/capability claims; no actual nonconsensual organizational victim or autonomous escape is identified. Nasdaq syndicates vendor release rather than independently corroborating an incident. Source checks: read: https://www.nasdaq.com/press-release/radware-unveils-zombieagent-newly-discovered-zero-click-ai-agent-vulnerability. Read Radware-authored GlobeNewswire release syndicated on Nasdaq. Vulnerability disclosure/testing with potential propagation claims; not independent reporting or an identified nonconsensual victim.
Evidence source
A review decision applies only to the named scope. Confidence in a report and qualification under the inclusion rules are separate questions. Review records do not add incidents to the globe or leaderboard.