Report
AI privacy flags didn’t stop prompt theft, researcher says
Ars Technica reports that HD Moore used nearly two million bait-filled prompts to trace data leaks.
TLDR
Security researcher HD Moore found evidence of at least three compromised AI providers, resellers passing along prompts, and providers ignoring privacy settings, Ars Technica reports. His canary credentials were bait, but attackers used the AWS keys in tens of thousands of API calls. Moore recommends self-hosting where possible or dealing directly with a first-tier provider, while warning that going direct still doesn’t guarantee privacy.
Combined views
9.1K
7 Sources, first seen ago

Comments