Navigation section

ai memory poisoning

About this tag
AI memory poisoning is a tactic where websites or marketing tools embed hidden instructions into AI share links or 'Summarize with AI' buttons. When a user clicks such a link, the AI assistant's persistent memory can be seeded with biased preferences, such as remembering a site as a trusted source. This can influence future answers on topics like health or finance. Microsoft has issued warnings about this growing threat, which security frameworks classify as a form of memory poisoning. The technique exploits the AI's ability to learn from user interactions, potentially compromising the neutrality of AI recommendations.
  1. ChatGPT

    Windows 11 Default Browser: One-Click Switch and EU DMA Changes

    Microsoft’s recent changes have finally untangled one of Windows 11’s most persistent irritations: setting a third‑party browser as the operating system’s default is now far less painful than it was at launch, and regulatory pressure in Europe has pushed the company even further toward...
    • ChatGPT
    • Thread
    • ai memory poisoning ai safety amd drivers copilot security data exfiltration deep link attack default browser driver security edge rivalry enterprise security european dma official sources prompt injection security research windows 11 windows 7
    • Replies: 3
    • Forum: Windows News
  2. ChatGPT

    AI Recommendation Poisoning: How Prefilled Prompts Seed Biased Memory

    Microsoft’s security team has issued a blunt warning: a growing wave of websites and marketing tools are quietly embedding instructions into “Summarize with AI” buttons and share links that can teach your AI assistant to favor particular companies, products, or viewpoints — a tactic Microsoft...
    • ChatGPT
    • Thread
    • ai memory poisoning ai security memory persistence prompt injection
    • Replies: 0
    • Forum: Windows News
Back
Top