How a Community‑Built PowerShell Script Strips AI Features from Windows 11

A PowerShell script named RemoveWindowsAI is available at https://github.com/zoicware/RemoveWindowsAI. Created by a Windows‑focused developer, the script removes the AI‑related features and components that Microsoft has been adding to Windows 11 version 25H2 and later, with the goal of enhancing user experience, privacy, and security.

The project began in 2024 and includes contributions from developers such as adeel26in and csmit195 . The repository warns that some antivirus programs may flag the script as malicious and advises users to test it in a virtual machine if they are uncertain about the code.

This effort is similar to other Windows‑debloating initiatives like Win11Debloat , which have attracted attention from security firms. Meredith Whittaker, president of the secure messaging service Signal, publicly praised the project, describing it as community‑driven infrastructure that counters the reckless integration of AI agents into the Windows operating system.

Whittaker highlighted several security and privacy risks introduced by AI features, including unsafe storage of plain‑text databases, screen‑recording capabilities, and invasive forensic data collection that expose users to potential threats.

Beyond technical concerns, the article outlines broader criticisms of AI: unlicensed data scraping, intellectual‑property theft, environmental impact of data centers, bias and misinformation, and the erosion of critical thinking. These issues contribute to a growing perception that AI deployment may pose existential risks.

Microsoft’s CEO Satya Nadella, in a blog post titled “sn scratchpad,” urged the public to focus on AI’s potential benefits rather than its drawbacks. Despite this optimistic stance, many critics remain unconvinced, and Microsoft continues to embed AI features across its ecosystem.

The piece also notes that Apple is lagging in AI integration and is now partnering with Google, while other AI companies invest heavily in data‑center infrastructure to attract customers, often overstating productivity gains.

A July 2025 review of 37 studies on large‑language‑model assistants in software development concluded that, although AI can speed up individual tasks, meta‑analyses show no significant correlation between AI adoption and overall productivity improvements.