Your unpublished ideas in chatbots could leak—experts on how to reduce the risk
_akpiper · x · 2026-10-22
A Science news report examines an overlooked risk: ideas, research directions, and business plans typed into chatbots or coding agents could end up used by the model's builder—or surfaced a year later to a rival who asks the right questions.
- Experts note no setting makes a chatbot fully private
- The piece outlines practical steps—around data retention policies and enterprise configurations—that can make leaks much less likely
- Particularly relevant for researchers, developers, and founders whose AI conversations are themselves an information exposure surface
More from Safety
- 13 engineers found 2 years' worth of severe bugs in a quarter with 5 simple LLM prompts — dr_cintas · 2026-10-22
- DeepSeek agent swarm reportedly patching vulnerable PHP boxes before attackers — burny_tech · 2026-10-22
- Banks brace for AI worst-case scenarios as rapid AI progress becomes a security nightmare — kimmonismus · 2026-10-22
- AI cyber offense vs defense is just the same old arms race, says security pro — ziv_ravid · 2026-10-22
- Open models will be the central pillar of AI cyber defense, argues security thread on PHP bugs — ziv_ravid · 2026-10-22
- Axios scoop: US banks detect unauthorized AI agents inside customer accounts — daniel_mac8 · 2026-10-22