2M-probe study exposes privacy leaks at AI providers
11/21/2026 — 11/21, 11:19·1 sources (all)·1 reports (all)
Key takeaways
- HD Moore, chief research officer at Dragos, sent nearly 2 million traceable canary requests to major AI providers to test for data leaks. At…
Background & analysis
On November 21, 2026, tech outlet Ars Technica reported that HD Moore, Chief Research Officer at the security firm Dragos, had sent close to 2 million requests carrying trackable decoys to multiple AI service providers in an effort to test whether submitted data was leaking out. The decoy secrets were deliberately planted by the researchers and were not drawn from real user sources. The goal was to watch for follow-up access from other programs or people, which would show whether data had escaped and through which part of the service chain it had travelled.
The findings showed signs of intrusion at no fewer than three providers. Attackers used stolen decoy keys to carry out tens of thousands of cloud API calls, seeking additional credentials, activating services and keeping access open. Some keys were used within hours of being submitted, and in certain cases a request was rejected before the model ever ran, yet its key was still taken. Risk also came from the layered reseller chain behind many AI applications. Requests are frequently routed through middlemen, so input can pass across multiple platforms and be decrypted at each step. The study found that some resellers logged requests and passed them on, and that private URLs inside requests tagged with zero data retention or data collection prohibited were still scraped days later. Moore recommends that sensitive workloads be run locally where possible, and that users who cannot self-host choose trusted providers directly and spell out their privacy terms.
AI-generated from 1 reports · updated 23 hours ago
- Heat index
- 49
- All sources
- 1
- All reports
- 1
- First seen
- 1 day ago
Reports on this story headlines open the original
HD Moore, chief research officer at security firm Dragos, sent nearly 2 million traceable decoy requests to multiple AI providers and found prompt theft, reseller leaks, and privacy settings that were not honored. At least three providers showed signs of compromise, with some keys used within hours, and URLs tied to "zero data retention" requests were still crawled days later. Moore recommends on-premise deployment for sensitive workloads or going directly to trusted providers.
IT之家AI score 77
Other stories people are talking about
- Heat index 1143RisingAnthropic Reaches $1.6 Trillion Market Value in First Trading Week After IPO21 sources
- Heat index 1011Anthropic Mathematician Says Claude Formalized the Generalized Riemann Hypothesis21 sources
- Heat index 622Anthropic IPO total raise reaches $93.5 billion after greenshoe exercise20 sources
- Heat index 596RisingThinking Machines Lab Open-Sources Inkling 2 Mini with 32GB Model Card Recommendation10 sources
- Heat index 490RisingSEC probes Nvidia accounting practices as shares post seven-week slide9 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

