TECH Signal 223
Scanning 7.6 Petabytes of HuggingFace Training Data for Secrets
↗
Secrets leaked into public AI training datasets can be discovered and exploited by anyone, creating a direct path to software supply chain attacks, cloud takeovers, and massive data exposure. The presence of live GitHub tokens with write access and Docker Hub push privileges means malicious actors could alter widely installed software.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
The scan found 349 live GitHub personal access tokens, including 223 with full repo write access and 130 capable of rewriting CI workflows, posing direct software supply chain risks.
Live credentials discovered included 8,557 GCP service-account keys across 3,811 projects, 8,594 working database logins, and one key that granted access to 393 GB of PII covering an estimated 3.7% of the global population.
Hugging Face's CTO contributed native storage-bucket scanning support to TruffleHog following responsible disclosure of the findings.
THE CLUSTER