API
< Back to all clusters
[TECHNOLOGY] · 2 sources

started · updated

HD Moore research reveals data leaks in AI supply chain

Security researcher HD Moore of Dragos conducted a study from July to November 2026 that uncovered significant data vulnerabilities within the AI service ecosystem. By embedding nearly 2 million trackable “canary” credentials and URLs into prompts, Moore tracked how sensitive information moved through AI routers, resellers, and hosting providers.

The investigation found that at least three AI hosting providers were compromised, allowing attackers to steal customer prompt data. In one instance during late October 2026, attackers reportedly exfiltrated approximately 200 unique AWS credentials. Additionally, the study identified instances where resellers logged prompts and failed to honor privacy settings, such as “Zero Data Retention” flags. The findings suggest that while prompts are encrypted during transit, they are decrypted at various stages within the inference network, exposing them to potential theft or unauthorized logging.

Entities

AWS · Dragos · HD Moore · OpenRouter · Vercel