AI & RoboticsMore in AI & Robotics→
NewsGuard audit finds Mistral’s Le Chat vulnerable to Iran-war disinformation prompts
Key Takeaways
- NewsGuard tested ten false Iran-war claims on Mistral’s Le Chat.
- The chatbot’s error rate reportedly rose from 10 percent on neutral prompts to 60 percent on leading ones and 80 percent on malicious prompts.
- The audit highlights how adversarial prompt framing can sharply degrade chatbot reliability.
DE
DT Editorial Team··via the-decoder.com