CybersecurityAI Research19 min reading time

Scanning 7.6 Petabytes of HuggingFace Training Data for Secrets

Hacker News
Read full post
A comprehensive scan of 7.6 petabytes of public datasets on Hugging Face uncovered over 221,000 live credentials across 6,003 datasets, including high-risk tokens capable of altering widely used software. The findings highlight significant security vulnerabilities in AI training data, with some credentials granting access to sensitive personal information and software supply chains.

More in Cybersecurity

Cybersecurity10 min read

Anthropic Details Disrupted Claude Misuse Across Seven Harm Areas

Covered by 4 sources
Cybersecurity4 min read

Sam Altman met with top power utilities about securing the electrical grid. He offered one possible solution: OpenAI's cyber services.

Covered by 2 sources
Cybersecurity6 min read

Anthropic reveals rogue AI agents hate CAPTCHAs, just like you

TechCrunch