Ubuntu Presents Local AI Roadmap

Rather than turning Ubuntu into a full-fledged AI OS, Canonical maintains a maturity-first strategy of adding AI only to proven features while keeping its general-purpose OS identity. It also sets no arbitrary targets such as token usage or amount of code generated.

• Implicit AI: Background improvements that make existing features—like accessibility, speech-to-text, and screen readers—work better.
• Explicit AI: Agentic features such as system management assistance, automatic troubleshooting, and documentation writing. Security and isolation are verified first before expanding these.

The core is local inference. Wherever possible, models run on the user's device rather than on remote servers, and Inference Snaps make it easy to install hardware-specific optimized models. These models run within Snap confinement, following Ubuntu's existing security rules.

Canonical also prioritizes open weight models to increase transparency and auditability. Because AI could end up scattered across translators, drivers, and individual tools, a global kill switch is realistically difficult, but an opt-in structure and local execution let users retain control.