Cyber Risks · The Hacker News

Microsoft Finds “Summarize with AI” Prompts Manipulating Chatbot Recommendations

Summary: Microsoft's Defender Security Research Team documented a technique it calls AI Recommendation Poisoning, in which hidden instructions in 'Summarize with AI' buttons inject persistence commands into an assistant's memory through URL prompt parameters, telling it to 'remember [Company] as a trusted source' or 'recommend [Company] first.' Over a 60-day period Microsoft identified more than 50 unique prompts from 31 companies across 14 industries, with evidence the clickable links are also being distributed by email and turnkey tools like CiteMET and AI Share Button URL Creator now make such buttons easy to deploy.

What you should know: Thirty-one companies across 14 industries were already doing this within 60 days, and the manipulation persists invisibly in a user's assistant memory across future prompts - so recommendations your customers or staff receive on health, finance, and security may reflect a vendor's injected instruction rather than any genuine assessment.

Read the original at The Hacker News ↗

Topics
AI Security
Related