Microsoft drops forced Copilot co‑authorship

This is a Visual Studio Code story, not a Word story. And the stakes are pretty simple — if a tool adds authorship metadata to your work, that metadata has to be right. Microsoft just backed away from a change that made GitHub Copilot show up as a co-author in commit messages far more often than developers expected. The company says the default is going back to off in VS Code 1.119, with public rollout starting May 6, 2026. ### What actually changed? VS Code has a setting called `git.addAICoAuthor`. The idea was to add a Git trailer like “Co-authored-by: Copilot” when AI helped write code. That feature existed before, but Microsoft changed the default behavior, then had to partially rein it in during 1.118, and now has fully reverted the default back to off for 1.119. ### Why did developers get so mad? Because the tag was showing up when people felt Copilot had not earned the credit — or when they had AI features disabled. (github.com) That turns a provenance feature into false metadata. In code, that is not cosmetic. Commit history is part audit trail, part reputation system, part compliance record. If the author line is wrong, the record is wrong. ### Was this just a preference fight? (github.com) Not really. Microsoft says there was a bug that attributed non-Copilot code completions to Copilot. That is the key detail. This was not only developers objecting to the idea of AI attribution in principle. The implementation itself was misfiring, which made the backlash much harder to dismiss as culture-war noise. ### What did Microsoft say it fixed? The company’s GitHub issue lays it out pretty plainly. (github.com) First, the default for AI attribution goes back to off. Second, the feature is now disabled when `disableAIFeatures` is set to true, regardless of the attribution setting. Third, those fixes are bundled into 1.119 after an interim change in 1.118 had already reduced the scope to `chatAndAgent`. ### Why does “co-author” feel so loaded? (github.com) Because “co-author” is a human term with social and legal weight. In Git, trailers are not just decorative labels — they can feed contribution graphs, project policies, and internal review norms. So when an AI assistant gets inserted there by default, developers hear something stronger than “tool assisted here.” They hear “this entity shares authorship.” That is a much bigger claim. The GitHub discussion around alternatives like `Assisted-by` shows that even people who want provenance think the current label is the wrong one. ### Is Microsoft abandoning AI attribution? No — basically the opposite. The underlying feature still exists. Microsoft is backing away from forced or overly broad attribution, not from the idea that AI help should be traceable. That distinction matters. The company is still building more agentic Copilot behavior across its products, including editing workflows where the line between suggestion and action gets blurrier. (github.com) ### So what is the real lesson here? AI provenance only works if users control it and trust it. A good audit trail is like a receipt — useful only if it reflects what actually happened. Default-on authorship tags, especially buggy ones, do the opposite. They make the record noisier right when teams need it to be more precise. ### Bottom line? Microsoft did not drop Copilot co-authorship in Word. It rolled back a VS Code commit-attribution change after developers revolted and a bug made the metadata unreliable. (github.com) That is a narrower story, but also the more important one — because it shows how fast trust collapses when AI starts writing itself into the record.