← Back to Model Beat
Research·13h ago·all news from November 22, 2026

Security researchers show how cheap AI inference providers steal keys and snoop on users

Security researchers at Dragos analyzed two million fake AI requests to identify vulnerabilities among third-party inference providers. The investigation revealed that several providers were actively stealing AWS keys and intercepting private user prompts. This discovery highlights significant supply chain risks for companies using smaller, secondary AI platforms, as these services can monitor or manipulate data transmissions. Following the findings, OpenRouter removed several providers from its marketplace.

Covered by 2 sources

Related stories

ResearchBytedance argues most AI training phases are unnecessary and backs it up with a competitive open modelNov 21ResearchResearcher plants 2m canaries into the AI supply chain. A privacy disaster surfaced.Nov 20 · 3 sourcesResearchNvidia researchers argue huge AI context windows are a dead endNov 21ResearchThis smell startup says its AI nose is as good as a human’sNov 18 · 2 sources