The Document Foundation (TDF) repeats that LibreOffice has no built-in AI features and that its core developers do not use language models. Code that comes in from outside follows different rules, and some contributors have openly said they used LLM-based coding assistants.
Three commits that cite AI
A user on Mastodon, who goes by argv[-1], found three LibreOffice commits from this year that acknowledge the use of LLM-based coding assistants. Two date from April and credit Claude with helping identify the fixes. The third, from September, states that the BitmapTransferable class was generated with GitHub Copilot in VS Code, while the copyButtonHandler method was written by hand with help from an LLM-based code-completion assistant.
Image: wiki.documentfoundation.org — https://wiki.documentfoundation.org/Development/GSoc
The fact is hardly surprising: almost no major open source project is now free of LLM-generated code. The issue is the public message. The project presents itself as AI-free, and that has become one of its selling points. In September TDF published a post declaring the suite free of AI features, and on October 21 it announced download growth, which it said was difficult to attribute to any single cause.
TDF's position
The foundation's reply is firm. On Mastodon, TDF says no core developers use LLMs and that the project relies on contributions from the wider community without explicitly listing forbidden tools. Contributors must disclose any LLM use, and the foundation generally discourages the practice. Every submission is reviewed for code quality and for whether the author understands the code. Students are prohibited from using LLMs.
Spokesperson Italo Vignoli, asked by The Register, pointed to the public statements on LibreOffice 26.8 and said they concerned user features, not development.
Image: libreoffice.org — https://www.libreoffice.org/download/?type
A ban that looks inconsistent
The distinction between user features and development does not convince everyone. Some LibreOffice users are unhappy, and an Argentinian developer compared the situation to a bakery promising no asbestos in its frosting while using it in the filling. Another comment, quoted in the story, finds it odd to ban all LLM use for commit messages while allowing it for the actual code.
The matter also touches Collabora, LibreOffice's former development partner, whose CODE web suite is based on LibreOffice code and which in June pushed AI integration more openly. For TDF, the no-AI line also marks a difference from that path.
The episode shows how a project can state a strong principle in public messaging while applying a more nuanced one in daily work, with transparency and human review in place of a ban. Whether that distinction holds up as AI-assisted contributions grow more common remains to be seen.
No comments yet. Be the first!