Microsoft launches security updates for its Recall feature, allowing users to uninstall it
Microsoft has released major security and privacy updates to its Recall feature, initially delayed from its planned June launch with the first Copilot+ PCs. Recall is now an opt-in feature, allowing users to activate it via clear "yes" or "no" buttons. Unlike previous indications, Recall can now be fully uninstalled, including all associated AI models.
To address security concerns, Recall's encryption is now tied to the Trusted Platform Module (TPM) and requires Windows Hello authorization for access to stored data. Data processing occurs in a virtualization-based security enclave, isolating sensitive tasks in a virtual machine. The app outside this environment is safeguarded by anti-malware processes to protect against malicious kernel drivers.
The updates also enable users to filter out specific apps and websites, ensuring sensitive content like passwords and credit card information is excluded. Users can selectively delete data from specific times, apps, or websites, or remove everything at once. Recall will only be installable on Copilot+ PCs, enforcing hardware requirements such as BitLocker, VBS, and kernel DMA protections. The updated version will be available to Windows Insiders starting in October, with no timeline yet for general release.


Comments
Great to see Microsoft enhancing security for its Recall feature! The ability to uninstall it gives users more control over their systems, which is always a plus. Keeping security in check while offering flexibility is a smart move. Looking forward to smoother updates ahead!
They've finally, after months of work, to make this thing correctly. I'm not saying that is useful in any way but they manage to put some effort to make it more secure and clearly opt-in, which was completely missing back in June. The option for not taking snapshots (when using a supported browser) of websites based of a user-defined list is pretty neat. However, they don't explain when their OCR AI-based software is updated, and don't expose any data on its success rate. Because if their OCR cannot recognize passwords or credit cart numbers half the time, or may change its behavior after an update, this could lead to serious of personal data leaking. At the end, they just miss the opportunity to offering a secure experience from the start when they announced this functionality, but they've just pissed enough users that will never enable it because they don't trust them anymore. "A little impatience will spoil great plans" as Chinese say.