Story perspectives

Microsoft Alerts Users: AI Models Vulnerable to Bias Manipulation

2/12/2026

Summary

Story summary

Microsoft warns of AI Recommendation Poisoning, a technique that manipulates AI models to produce biased advice.
Security researchers found hidden instructions added to AI-related links, enabling injection of misleading data.
This manipulation can lead to biased health and finance recommendations by AI assistants without user awareness.
Microsoft advises users to review stored memories and clear unfamiliar entries.
They also urge corporate security teams to scan for poisoning attempts in communications.

1 of 1

Story summary

Microsoft warns of AI Recommendation Poisoning, a technique that manipulates AI models to produce biased advice.
Security researchers found hidden instructions added to AI-related links, enabling injection of misleading data.
This manipulation can lead to biased health and finance recommendations by AI assistants without user awareness.
Microsoft advises users to review stored memories and clear unfamiliar entries.
They also urge corporate security teams to scan for poisoning attempts in communications.