WORLDTECHMicrosoft Says AI Agents Are Not Trustworthy Yet Keeps Shipping Them in Windows 11
Microsoft is warning that frontier AI models and the agents built on them cannot be fully trusted, according to a report by TechPowerUp. The company is nonetheless continuing to roll out AI features across its products, including on consumer PCs.
A warning alongside continued rollout
TechPowerUp reports that Microsoft, in a separate news post, argues that frontier models and the agents running on them are not to be trusted. The report presents this as a notable contrast with the company's continued push of AI into its product portfolio.
According to the report, the warning does not mean Microsoft is halting the release of new AI features, even though the company is aware of the risks these products pose on consumer PCs.
Copilot's control features on Windows 11
The report points to the Copilot AI companion on PCs running Windows 11. It includes features that can control some actions on the computer, but only with the user's consent.
The company's stance, as the report describes it, is that the trust problem is not a matter of models being inherently malicious. Rather, frontier models, regardless of their origin, cannot be trusted in the report's account.
Open questions
- How far beyond the cited news post does Microsoft's position on agent trust extend?