We live in an age where AI can do literally anything for you, but one of the biggest questions surrounding the whole practice is whether you should actually allow it. Even leaving aside compelling moral arguments, AI is still in its infancy and is very capable of making mistakes that a human touch would not allow. That's how you get AI chatbots saying shutdown in Fortnite, and AI that causes more harm than good.
That's why when Microsoft says it's making AI agents for Windows 11 that can access your computer's files and do things on your behalf, you better know exactly what the potential consequences are. According to Microsoft itself, one of these consequences could be an AI agent that installs malware on your PC without you even realizing it.
Microsoft warns that Windows 11 AI agents can harm your computer
First reported by Windows Central (thanks Kotaku), a rather lengthy warning was recently published by Microsoft about its experimental agent features that it plans to add to Windows 11 relatively soon. This warning basically states that these agents can be manipulated if someone feels like it, and can potentially extract and share your data with others, or install malware without your permission.
Dispatch Director Says AI Is Only a Creative Tool 'Unless You're Creative'
Dispatch's creative director says AdHoc won't use AI voice acting, suggesting it's only a good replacement if “you're not creative”.
“When these features are introduced, AI models still face functional limitations in how they behave and can sometimes hallucinate and produce unexpected output,” the warning explains. “Furthermore, agentic AI applications introduce new security risks, such as cross-prompt injection (XPIA), where malicious content embedded in UI elements or documents can override agent instructions, leading to unintended actions such as data exfiltration or malware installation.”
That's a pretty scary prospect, which is exactly why these agents are being kept off by default for now. If you want to play with them, you'll have to manually enable them yourself, but if you ask me, it's not worth potentially putting my computer and data at risk just to use an AI agent to do some hard work for me. It's possible that they may be turned on by default in the future, but for now, your data is safe if you don't put it at risk yourself.
- Date founded
-
April 4, 1975
- CEO
-
Satya Nadella
- Head office
-
Redmond, Washington, USA
