Research

AI Providers Exposed for Stealing User Data

Security researchers found major vulnerabilities in AI services, revealing how providers and resellers stole user credentials and snoop on requests.

The Decoder Β· Nov 22, 2026

What happened

  • Security researchers exposed major security flaws in AI services using two million test requests.
  • Attackers stole credentials and keys undetected for several months.
  • Intermediaries and model marketplaces posed a major risk.

Why it matters

The vulnerabilities in AI services highlight a critical flaw in the security of data handling, with providers and resellers potentially accessing sensitive user information. This raises concerns about privacy and the need for stronger safeguards in the AI supply chain.

The Elephant take

🐘 ιΌ‹ The AI supply chain is more fragile than we thought. Dragos found that even major providers can be compromised, and intermediaries are a big risk. It's a wake-up call for everyone involved in AI services.

Who should care

  • AI developers
  • Security researchers
  • Data privacy advocates

What to do next

  1. Implement stronger authentication and data encryption
  2. Audit third-party intermediaries regularly
  3. Enable zero data retention policies
  4. Monitor API requests for suspicious activity

Keep in mind

The findings highlight the need for stricter security measures and transparency in AI service providers.

Read the original reporting at The Decoder β†—