A security researcher seeded roughly two million uniquely trackable secrets into prompts sent through AI service networks to trace where sensitive data ends up, and the experiment found leaks at multiple points in the chain. The test matters because many AI apps rely on intermediary routers and gateways that developers do not directly control, so privacy promises made at the app level may not survive the handoffs.
Source: https://arstechnica.com/security/2026/11/a-leaky-ai-supply-chain-is-exposing-secret-user-data-at-an-alarming-rate/
Feed publication: 2026-11-20 20:59 UTC
Details
According to the excerpt, researcher HD Moore built fake but functional credentials, including AWS keys tied to empty accounts, SSH passwords, and unique URLs, DNS names and private git references. Each secret appeared in exactly one prompt sent to a single provider, letting him attribute any later misuse to a specific link. The reported findings include at least three AI providers compromised by attackers who siphoned prompts, resellers in the inference marketplace that logged prompts and passed them on, and providers ignoring privacy flags such as Zero Data Retention or no-data-collection settings.
Source: https://arstechnica.com/security/2026/11/a-leaky-ai-supply-chain-is-exposing-secret-user-data-at-an-alarming-rate/
What remains uncertain
The supplied material is a partial article excerpt, so the full methodology, the named providers and resellers, and any responses from the companies involved are not available here. These are partial article and document excerpts, not independent verification.
Primary source: https://arstechnica.com/security/2026/11/a-leaky-ai-supply-chain-is-exposing-secret-user-data-at-an-alarming-rate/
Discussion
If you use AI apps that route requests through third-party gateways, do you know what data-retention settings actually apply along that chain?
Написать Join — чтобы участвовать в обсуждении.