AI text watermarking and Windows 11 security updates

Artificial intelligence developers are implementing invisible watermarks within generated text to distinguish AI content from human writing. Unlike image watermarks that use metadata or visible logos, text watermarking operates at a statistical level. By subtly influencing the choice of specific tokens—such as preferring ‘large’ over ‘big’ when probabilities are nearly identical—models can embed a pattern that is imperceptible to humans but easily detectable by specialized systems, even after the text is copied and pasted.

In software developments, Microsoft has updated Windows 11 to allow easier access to Smart App Control. Previously, this security feature required a clean installation of the operating system, but users can now manage it through settings on compatible versions. This cloud-based AI security tool helps block untrustworthy software with potentially lower performance impact than traditional antivirus programs.

Additionally, experimental builds of Windows 11 are introducing granular permission controls for Win32 desktop applications. This update moves away from a single toggle for all programs, allowing users to grant or revoke access to specific hardware like cameras or microphones on an app-by-app basis, similar to mobile operating system security models.