Report
Researcher says AI prompt tripwires exposed breaches and broken privacy promises
HD Moore says nearly two million test prompts exposed stolen credentials and ignored data-retention settings.
TLDR
Security researcher HD Moore seeded nearly two million AI prompts with trackable secrets, Ars Technica reports. He says the test exposed at least three compromised providers, resellers leaking prompts, and services ignoring privacy settings. Stolen bait AWS keys generated tens of thousands of API calls. Even some requests marked for zero data retention had secrets stolen, Moore says.
Combined views
81.6K
1 Source, first seen ago

Comments