Free Consultation
Home Incident Radar Model data poisoning (PoisonGPT)
AI incident

Model data poisoning (PoisonGPT)

Research (Mithril) · 2023

Organisation
Research (Mithril)
Year
2023
Failure mode
Data / model poisoning
Severity
critical
Source
Mithril Security, 2023

What happened

A tampered model spread targeted misinformation while passing standard benchmarks.

What would have prevented it

Data & model provenance + supply-chain integrity

Requirements this maps to

Editorial cross-reference from our prevention note — it is not a finding that any law was breached.