14-09-2026
PuzzleMask: Abusing Plain Prose as a Covert AI Attack Vector
https://research.checkpoint.com/2026/puzzlemask-abusing-plain-prose-as-a-covert-ai-attack-vector/
Report completeness: Low
Threats:
Puzzlemask_technique
Injecagent_tool
Redagent_tool
Victims:
Llm gatekeeper and target pipelines, Llm applications
TTPs:
Tactics: 2
Technics: 0
ChatGPT TTPs:
do not use without manual checkT1027, T1059.006, T1140
IOCs:
File: 1
Soft:
gatekeeper, gatekeepers, gpt-5, OpenAI, Anthropic, Fable 5, atekeeper, ekeeper
Algorithms:
base64
Languages:
python
