Canonical is funding a PhD to find out if AI can turn millions of lines of C into Rust

Key points

• Canonical is co-funding a three-year PhD at the University of Bristol to study whether AI can automatically translate large C codebases into the Rust programming language.
• The project is co-sponsored by UK Research and Innovation, a public body backed by the UK's Department for Business, Innovation, Science and Trade.
• Ubuntu engineering VP Jon Seager announced the project on Ubuntu's Discourse forum in 2025.
• The research will focus on two real tools: snap-confine and AppArmor, both written in C.
• This is exploratory research, not a plan to rewrite Ubuntu itself.

Canonical, the company behind Ubuntu, is putting money behind a question the software world hasn't settled: can an AI read a huge pile of old C code and rewrite it in Rust automatically?

C is a programming language that's powered software for decades. It's fast, but notoriously easy to write in ways that leave dangerous security holes. Rust is a newer language designed to block that whole class of holes from the start. Converting one to the other by hand is painfully slow.

What exactly is Canonical paying for?

A PhD student at the University of Bristol's Programming Languages Research Group will spend three years investigating whether a large language model (an LLM, the technology behind tools like ChatGPT) can reliably do that translation. Professor Meng Wang and Dr Cristina David will supervise the student alongside Seager.

The plan isn't to paste an entire operating system into a chatbot. First, a giant codebase (potentially hundreds of thousands of lines) gets broken into smaller chunks. Then an AI rewrites each chunk in Rust that's meant to be safe, correct and readable by a human maintainer. The two test subjects are snap-confine and AppArmor, both security-related components already used in Ubuntu.

Think of a 30-year-old filing cabinet stuffed with handwritten notes, some in a language only three people ever spoke. This project asks whether AI can read it all, understand the logic behind every scrawled margin note, and type it up cleanly in a modern format without losing anything. Mature codebases carry decades of knowledge baked silently into the code itself: patches for bugs nobody documented, workarounds for edge cases that only appeared in production. An AI translation, done carefully, might preserve them where a human rewrite from scratch could miss them entirely.

It's worth pairing this with our 30 July story on Chrome patching security holes twice a week with AI-found bugs, because both cases point at the same underlying shift: AI isn't just writing new code, it's being aimed at the vast stock of old code that nobody wants to touch by hand.

Will this actually work?

Honestly, nobody knows. That's the point.

Existing automated translation tools tend to produce Rust code that still behaves like C under the hood, leaning on "unsafe" operations and awkward patterns a Rust developer would immediately want to fix. Seager's proposal, which runs to just over 1,000 words, addresses this directly and is worth reading before dismissing the idea.

The harder challenge may not be the translation itself but the decomposition step: figuring out how to slice a sprawling codebase into pieces small enough for an AI to handle without losing the connections between them. That's an unsolved problem, and three years of research may confirm it's genuinely very hard.

One thing that could quietly limit the AppArmor side of the work: openSUSE 16 replaced AppArmor with SELinux, and enterprise Linux has largely consolidated around SELinux too. A hardened Rust version of AppArmor would still matter for Debian and several smaller distributions, but its audience is narrower than it once was.

What does this mean for ordinary Ubuntu users?

Nothing changes now. This is research, not a release schedule. If it succeeds, the long-term payoff would be software that's harder to exploit, arriving faster than if developers rewrote everything by hand. Real public money is involved, which means real accountability.

Common questions

Is Canonical planning to rewrite all of Ubuntu in Rust?

No. This project investigates whether AI-assisted translation is even feasible on large codebases. A full Ubuntu rewrite isn't on the table.

What is Rust and why does it matter?

Rust is a programming language designed to eliminate a category of security vulnerabilities common in older languages like C. More Rust in core software generally means fewer exploitable bugs.

How long before this affects any real software?

At minimum, three years for the research itself. Any practical adoption based on the findings would come after that.