2M canary-token prompts reveal AI provider breaches and prompt theft
WeldPond · x · 2026-11-20
Dragos CSO HD Moore's team seeded 2 million prompts containing unique canary tokens and URLs into widely used AI services over five months (July–November 2026), then tracked where those secrets surfaced. Findings: evidence that data from at least three AI providers was compromised, with seeded credentials used against their test systems and canary AWS accounts; multiple resellers on an AI inference marketplace viewing and retaining customer prompts via standard logging; and privacy settings/retention policies being ignored. The writeup explains how canary tokens detect prompt leaks and how to protect your AI supply chain.
Related event: 2 Million Canary Tokens Expose AI Supply Chain Data Leaks(2 posts)→
More from Infra
- Wedbush raises Nvidia price target to $293, says leadership risks already priced in — sudoraohacker · 2026-11-20
- DeepMind essay maps four paths to superintelligence: effective compute grows 10x per year — udmrzn · 2026-11-20
- AtlasKV: billion-scale knowledge graphs in 20GB VRAM, no retriever or vector DB needed — udmrzn · 2026-11-20
- Open weights jump from 7.3% to 30.9% of Databricks-hosted tokens in four months — AIFlow_ML · 2026-11-20
- 7 web search APIs with MCP servers for Claude and Cursor, benchmarked — JeremyCMorgan · 2026-11-20
- Baseten ships deferred tool loading: agents run up to 62% cheaper and 26% faster — baseten · 2026-11-20