You don't have to have it if you don't want it.
In an era when artificial intelligence has become something tech companies impose rather than offer, Canonical has chosen a quieter path with Ubuntu: making AI a door users may open, not one they are pushed through. The company's roadmap centers on opt-in integration, local inference, and a genuine removal mechanism — not a buried toggle — for those who want no part of it. This is less a technical announcement than a philosophical one, a statement that user autonomy and technological progress need not be in conflict.
- While Microsoft, Apple, and others have embedded AI deeper into their systems with little room for refusal, Canonical is moving in the opposite direction — treating AI as an invitation, not a default.
- The tension is real: the tech industry has normalized AI that arrives uninvited, running in the background, processing data, and reshaping user experience whether consent was given or not.
- Canonical's answer is structural — users can remove the snap packages that deliver AI features entirely, creating a kill switch that is architectural rather than cosmetic.
- Local inference capabilities mean that for users who do opt in, data need not leave the device, directly addressing the privacy anxieties that have made many wary of AI adoption.
- The announcement lands as both a product strategy and an implicit challenge to the industry: there may be a significant market of users who simply want to choose.
Canonical, the company behind Ubuntu Linux, has outlined an AI strategy built on a premise that has become rare in the tech industry: users who don't want AI features don't have to have them. Rather than baking AI into the default experience, the company is making it opt-in — delivered through snap packages that can be removed entirely, leaving no hidden processes or forced updates behind.
This stands in deliberate contrast to the direction taken by Microsoft, Apple, and others, who have moved toward deep, often unavoidable AI integration. Canonical is framing its approach as a matter of user autonomy — a recognition that not everyone wants language models running on their system or their data participating in cloud-based services.
The technical design reflects that philosophy. Local inference capabilities mean AI processing can happen directly on a user's machine, keeping data off remote servers. The company is also developing what it calls agentic system tools — task-specific AI features rather than general-purpose assistants — designed to feel purposeful rather than intrusive.
What distinguishes this announcement is the choice it represents. Canonical could have followed the industry toward ubiquitous AI integration and made user control the exception. Instead, it inverted that hierarchy, betting that users who want AI will seek it out, and that those who don't deserve equal respect for that preference.
Whether this approach reshapes how the broader industry thinks about AI adoption remains uncertain. If the market rewards it, the model could shift the conversation. If not, Ubuntu's careful restraint may simply stand as a record of a different road — one that was available, and taken.
Canonical, the company behind Ubuntu Linux, has sketched out a vision for artificial intelligence that starts with a premise many users have come to crave: you don't have to have it if you don't want it. The plan, laid out recently, centers on optional AI tools rather than the kind of mandatory integration that has become standard across much of the tech industry. The mechanism is straightforward—users who wish to avoid AI features entirely can simply remove snaps, the containerized software packages that would deliver these capabilities. No hidden toggles, no persistent background processes, no forced updates that slip AI into your system whether you asked for it or not.
This approach stands in deliberate contrast to what competitors have been doing. Microsoft, Apple, and others have moved toward baking AI deeper into their operating systems and applications, often with limited user control over whether the features activate. Canonical's strategy instead treats AI as something users should actively choose to adopt, not something that arrives as part of the default experience. The company is framing this as a matter of respect for user autonomy—a recognition that not everyone wants their system running language models, making inferences about their data, or participating in cloud-based AI services.
The technical architecture reflects this philosophy. Rather than relying solely on cloud-based AI services, Canonical's roadmap includes local inference capabilities, meaning some AI processing would happen directly on a user's machine rather than being sent to remote servers. This addresses privacy concerns that have shadowed AI adoption: if the computation happens locally, the data doesn't leave your device. The company is also planning what it calls agentic system tools—AI features designed to perform specific tasks within the operating system itself, rather than general-purpose chatbots or assistants that might feel intrusive to users who simply want their computer to work.
What makes this announcement noteworthy is not that Canonical invented these ideas, but that it chose them as its primary direction when it could have chosen otherwise. The company had the option to follow the industry trend toward ubiquitous AI integration, to make AI the default and user control the exception. Instead, it inverted that hierarchy. Users will have to opt in. They will have a genuine kill switch—not a buried setting in a preferences menu, but the ability to remove the entire snap infrastructure that delivers these features. This is a bet that some users will want AI tools and will actively seek them out, while others will prefer their systems without them, and both groups deserve to have their preference respected.
The announcement has drawn comparisons to how Canonical has historically approached other contentious features. The company has a track record of listening to its user base when integration decisions prove unpopular, though it has also made choices that alienated segments of its community. This AI roadmap suggests the company learned something from those experiences—that forcing features on users, even well-intentioned ones, creates friction and resentment. By making AI optional from the start, Canonical is attempting to avoid that dynamic entirely.
What remains to be seen is whether this approach influences how other companies think about AI integration. The tech industry has largely treated AI as something that should be everywhere, integrated into every product, pushed to every user. Canonical's bet is that there's a market—perhaps a large one—for systems that let users decide. If that proves true, it could shift how the conversation around AI adoption happens across the industry. If it doesn't, Ubuntu's thoughtful approach may simply become a footnote, a road not taken by the companies that dominate consumer computing.
Notable Quotes
Canonical's approach to AI is refreshingly thoughtful— ZDNET
The Hearth Conversation Another angle on the story
Why does Canonical think users want to opt out of AI entirely? Isn't the whole industry moving toward integration?
The industry is moving that way, yes. But Canonical seems to be betting that forced integration creates backlash. They're saying: let people choose, and you'll build trust instead of resentment.
But doesn't that assume users understand what they're opting into? Most people don't read the fine print.
Fair point. But the kill switch—removing snaps—is so simple that even non-technical users can do it. It's not buried in settings. It's not a toggle you have to find. It's just: don't install the thing.
What about the local inference part? Why does that matter?
Because it means your data doesn't have to leave your machine. Cloud-based AI means Canonical or whoever runs the servers sees your data. Local inference keeps it private. That's a real difference for people who care about privacy.
Is this actually a competitive advantage, or just good PR?
Probably both. But the PR only works if the feature actually exists and works well. If Canonical ships broken local inference, users will notice. The advantage is real only if the execution is real.
What happens if Microsoft or Apple copy this approach?
Then Canonical loses the differentiation. But they've already won something: they've shown it's possible to do AI differently. That changes the conversation, even if others catch up.