Archive
Format: Research explained
-
How o3 helped a researcher find a Linux SMB memory-safety flaw
The vulnerability affected Linux ksmbd; OpenAI’s model was used during discovery.
-
How hidden instructions in documents can steer a multimodal AI agent
Trend Micro’s Pandora demonstration examines indirect prompt injection through content an agent is asked to process.
-
AI-agent security failures can come from application design rather than the framework
Unit 42 tested similar applications built with different frameworks and found recurring risks at tool and permission boundaries.
-
DeepSeek commentary highlights exposed infrastructure, not confirmed dark-web resale
A 2025 commentary uses DeepSeek’s database exposure to discuss downstream data risks. Its framing is not evidence of a confirmed resale.
-
Check Point describes faster vulnerability research with OpenAI Daybreak
A vendor case study describes faster defensive research. Its claims need to be distinguished from independently measured results.