AI Spies and Broken Features

It's been a troubling week for Microsoft, with a string of issues and warnings that raise serious concerns about the stability and security of Windows 11. The company's push for new AI-powered features, particularly under the term "agentic AI," has left both users and industry experts questioning the reliability of the operating system as a whole.

Recent warnings from Microsoft highlight how these AI agents could be vulnerable to tampering, potentially leading to your data being shared without permission or worse, malware being installed on your PC. Outlined in this source from The Gamer.

What is Agentic AI?

Agentic AI refers to autonomous systems capable of making decisions and performing tasks with minimal or no human intervention. This is a step beyond earlier technologies like Robotic Process Automation (RPA), which was used to automate rule-based, repetitive tasks rarely with anything more than a neural network to handle more complex but still rule-and-logic based decisions at its core functionality and is not reliant on the AI we know today. While RPAs were limited in scope and mostly functioned with fixed logic it is agentic AI that takes it further by enabling systems to respond dynamically to changes in the environment and make decisions on their own. This next level of automation can process results without any direct human input, which makes it a powerful tool for automating more complex tasks.

In short, agentic AI is a significant leap forward in artificial intelligence, and RPAs, enabling systems to perform intricate tasks without constant supervision, which while exciting for science, as we'll see, is potentially dangerous if not implemented carefully for everyone regardless.

Security Risks and Uncertain Future

Microsoft's warning about its experimental AI tools for Windows 11 has created quite a stir. These tools, intended to automate tasks across the operating system, come with serious security risks. The AI agents, still in development, are prone to "hallucinations" situations where the AI generates inaccurate or unexpected results. However, what's even more troubling is the possibility of cross-prompt injection (XPIA), this is a type of security flaw where malicious actors could manipulate the AI into performing harmful actions, like stealing data or installing malware!

Given these vulnerabilities, Microsoft has kept these AI tools disabled by default, requiring users to manually opt-in if they want to test them out. However, this raises an important question what happens if these AI features are eventually enabled by default?
The risk is high, especially considering the potential long-term implications of having agentic AI systems integrated deeply into the OS without fully addressing these security flaws.

Microsoft's Ongoing Patch Issues, Crashes, and Downtime

In addition to AI concerns, during the same period, Microsoft has faced several setbacks with its other services. Recently, Microsoft 365 suffered a major outage that prevented users from accessing files an issue that has unfortunately become all too familiar. Similarly, not too long ago, we witnessed Microsoft Azure go down entirely. But these are just two of their ongoing problems. The latest Patch Tuesday update has caused performance issues in games, prompting Nvidia to release an emergency hotfix driver. These types of performance issues are becoming increasingly common, with updates that are supposed to enhance system performance instead causing crashes, slowdowns, and instability.

This highlights the growing frustration with Windows 11 updates. Microsoft's broader effort to transform Windows into an "agentic" operating system has raised red flags, as more and more features are being rolled out before they are fully reliable, prompting some users to consider switching to Linux-based distributions or even moving to Apple, Microsoft's largest competitor, in the PC and OS market. For many users, these growing pains with Microsoft are becoming impossible to ignore, and the idea of integrating more automated, AI-driven systems into Windows 11 is starting to feel more like a risk than an advancement.

While Apple has its own AI ecosystem, they've taken a much more robust, privacy-centric approach. In recent months, Apple has even publicly argued against the Online Safety Act (OSA) and the UK government's attempts to monitor and scan user messages within the Apple ecosystem-reaffirming their commitment to user privacy and consumer care.

Apple products are notably more expensive for the average consumer than Microsoft's offerings, some are willing to bite the proverbial bullet in order to secure their privacy and ensure their data stays out of cloud-driven AI 'hands'. Instead, they're opting for Apple's on-device solutions like Siri and Core ML, which are baked directly into their AI ecosystem.

XAML Problems and Core Features Crashing

As if the security and performance issues weren't enough, Microsoft is also dealing with serious problems in core Windows 11 features. A July 2025 Patch Tuesday update (KB5062553) introduced issues with key components of Windows 11, including the Start Menu, Taskbar, File Explorer, and Windows Settings. These problems stem from the XAML framework, which powers the UI elements of Windows. As a result, core Shell components like Explorer.exe, StartMenuExperienceHost, and SystemSettings have been failing to load or crashing unexpectedly. Source for these issues can be found at Neowin.

The most frustrating part? These issues have been ongoing since July, but it took Microsoft until November to officially acknowledge them. And while the company has provided some temporary workarounds-such as restarting the Shell Infrastructure Host (SIHost.exe) service or using PowerShell commands to register the affected components, these fixes are far from ideal. For many users, the inconvenience of troubleshooting these core components just to get basic functionality back up and running is becoming increasingly unbearable.

Windows 11's Rocky Road Ahead

All these struggles paint a concerning picture for the future of Windows 11 and Microsoft. The promise of AI automation and powerful new features is undoubtedly exciting for some, but the reality is that these technologies are still in their infancy and fraught with security risks and in some most cases ethical dilema behind them. When combined with ongoing issues with core features and unreliable updates, it's clear that Windows 11 has a long way to go before it can be considered stable.

For now, it's probably best to keep those experimental AI features off in my opinion. With broken features, security vulnerabilities, and a sense that much of the OS is still in development, users may want to wait for a more stable version of Windows 11 before fully embracing these ambitious new updates. Microsoft has a lot of work to do - not just to fix bugs but to regain user trust.

Thank You ❤

If you ever need help, join the Discord or contact me directly.
Email: charlesbaron@tutanota.com