Microsoft's AI chatbot will 'recall' everything you do on its new PCs
Microsoft’s new “Recall” feature for Copilot+ Windows PCs, which continuously screenshots and indexes on-screen activity so an AI assistant can search your past actions, is raising sharp concerns about privacy, consent, and user control. Commenters fear the system will normalize pervasive tracking on personal computers, be hard to disable, and eventually tie into cloud services despite Microsoft’s claims that it runs locally. Many see it as part of a broader trend of ad- and data-driven design in Windows, prompting renewed interest in Linux, regulatory scrutiny under laws like GDPR, and calls for stronger ethical standards in software.
Overall reaction to Recall
- Dominant tone is distrust and frustration; many see Recall as the latest step in Windows becoming adware/spyware rather than a “personal” OS.
- Some argue this could be a genuinely powerful feature if fully local and user-controlled.
- A minority frames excitement around AI-assisted “photographic memory” and better retrieval of past work.
Privacy, consent, and surveillance
- Major concern: continuous screenshots of everything on screen effectively create a perfect surveillance log of users’ lives.
- Fear of misuse by employers (“bossware”), governments, law enforcement, or future regimes, including for retroactive punishment if laws change.
- Widespread skepticism that consent will be meaningful: patterns like “Yes / Not now” instead of “No,” repeated nags, and settings hidden or reset.
- Anxiety that Recall-like data will be used to train models and/or viewed by human labelers.
Local vs cloud and security
- Microsoft claims Recall runs on-device using the NPU, with storage quotas and on-disk snapshots.
- Some note it “has to” be local today for cost reasons, but worry it may shift to the cloud later.
- Others stress that even local storage is a huge security risk: a single compromise could expose months of highly sensitive activity.
Legal, regulatory, and antitrust angles
- EU’s GDPR and DMA are repeatedly cited as likely constraints; people expect region-specific limitations or pushback.
- Discussion of whistleblowing to regulators (EU + SEC) if companies knowingly violate privacy laws.
- Some wonder if OS-level usage tracking tied to an AI assistant could raise monopolistic behavior concerns.
Impact on OS and user choice
- Many say this accelerates their move to Linux or macOS; gaming on Linux (via Proton/Steam Deck) is cited as making migration feasible.
- Others predict three paths: embrace new AI-heavy Windows, stick with Windows 10 as long as possible, or rely on corporate IT to neuter features via policy.
- Tools like debloated Windows builds (e.g., AtlasOS) are suggested as partial mitigation; some think at that point you might as well switch OS.
Ethics, responsibility, and dark patterns
- Long subthread on who’s responsible for hostile UX (individual engineers vs “the machine” of incentives); no consensus.
- Calls both for regulation and for individual employees to refuse unethical work, but also recognition that economic pressure makes this hard.
Alternatives and similar ideas
- Comparisons to existing tools like Rewind (macOS) and to experimental “photographic memory” LLMs.
- Some ask for an open-source, fully local equivalent they can control themselves.