# AI made code free. Accountability is the scarce part now. > Source: > Published: 2026-07-16 00:00:00+00:00 The Signal · Issue 1 # AI made code free. Accountability is the scarce part now. Godot just told AI agents to stay out of its repo. The reason it gave is not that the code is bad. It is that no one is standing behind it, and that is the one thing you still get paid for. On June 30, the Godot game engine, one of the largest open-source projects on GitHub, published a contribution policy that reads like a door closing. Autonomous AI agents that open pull requests now trigger an automatic ban. Substantial AI-written code is refused, with the tool allowed only for menial help like completion, regex, and find-and-replace. If AI assisted, you disclose it. When you talk to a maintainer, a human writes the message. The line that matters is the reason. Godot’s maintainers did not say the output was low quality. They said **“AI cannot take responsibility,”** and that they cannot trust heavy AI users to understand their own submissions well enough to fix them later. Read that twice, because the objection is not to the code. It points at who is on the hook. This is where the week stops being a Godot story. Over the winter, cURL’s maintainer paused the project’s bug bounty after a wave of AI-generated vulnerability reports, most of them false, overwhelmed his reviewers, then reopened it in March once the volume eased. Earlier in the year, Mitchell Hashimoto put drive-by AI pull requests off-limits on Ghostty. A public tracker now lists open-source projects with AI-contribution rules, NetBSD, Gentoo, and Zig among them, and others have moved on their own: Ladybird stopped taking public pull requests this June, and Flathub restricted AI-generated submissions around the same time. The list has been growing for two years. Godot is not the start of something. It is the moment the pattern got big enough to argue about. The same week, Amazon’s CTO Werner Vogels made a parallel case in a Fortune interview. As AI writes more of the code, he argued, the work that matters moves to judgment: deciding whether the output should ship, and owning it once it does. He framed ownership in regulatory terms, that you cannot tell a regulator the model made the mistake, because a person still has to answer for the result. Godot was ruling on volunteer patches and Vogels was talking about payroll, and both arrived at the same standard: someone has to be accountable for the code. Here is the argument. Open source never actually ran on code. Code was always the cheap part, freely copyable and freely forkable. What made a contribution worth a maintainer’s time was the human attached to it: someone who understood the change and would come back to fix it when it broke. That attachment is what review is for, and it is what a pull request was really carrying. AI severed the human from the code. A model can now produce a patch that compiles and passes a first glance, with no one behind it who could tell you why line 40 is there. The code got free. The accountability went to zero. Maintainers are not banning AI because the output is bad. They are banning it because the output arrives orphaned, and reviewing orphaned code is unpaid work with nobody to hand the maintenance to. Godot said the quiet part out loud: reviewing AI submissions is demoralizing, because the feedback never mentors a future maintainer, it just gets absorbed by a machine. So the bans are a repricing. The commons is marking free code down to what it is now worth, close to nothing, and marking up the scarce thing, a human who understands the change and will stand behind it. ## What this means if you build alone You are on both sides of this line, and the second side is the one that pays. As a contributor, the free ride is ending. Shipping an AI patch you cannot explain into a project with a policy like Godot’s is now a fast way to get your account banned. If you want open-source standing, contribute where you understand the code well enough to defend it in review. That is a shorter list of pull requests and a far stronger one. As a builder, this is the clearest signal yet about where your margin is. The market is doing to your work exactly what Godot did to contributions: pricing generated output at zero and paying for the human who can vouch for it. Nobody will pay you to run a prompt they could run themselves. They will pay you to be the one who caught what the model got wrong and signed off on the result. On everything you ship, keep the receipts that prove a human was in the loop, because that is the part with a price on it. We made this case in [The $5,000 tell](/study/the-5000-dollar-tell/), about vendors quietly selling human review as the premium layer over their own AI. This week two very different rooms, a volunteer game engine and a trillion-dollar cloud, said it out loud in the same breath. The code is free now. Being accountable for it is not. One email, when there's something worth sending ## Get the research in your inbox. No fixed schedule, no filler. You get an email when we've tested something, run the numbers, or found a tool worth your time. Free. Double opt-in, unsubscribe in one click. Got a signal we missed? [Bring it to the forum ↗](https://community.okaneland.com) ## Sources & how we researched this [Godot contribution policy, June 30, 2026 (primary). Autonomous AI pull requests auto-banned, substantial AI code refused, disclosure required.](https://godotengine.org/article/contribution-policy-2026/)- The Register on the Godot policy, around July 1, 2026. - Werner Vogels on judgment and ownership as the engineer's value, Fortune, July 9, 2026. - cURL paused its bug bounty over AI slop and returned in March: Daniel Stenberg and The Register (winter 2026, context). - Ghostty on drive-by AI pull requests, Mitchell Hashimoto (early 2026, context). [The tracker of open-source AI-contribution policies (NetBSD, Gentoo, Zig and more). Ladybird ended public PRs in June 2026; Flathub restricted AI-generated submissions separately.](https://github.com/melissawm/open-source-ai-contribution-policies) The Signal is Okane Land's weekly read on the AI news that changes a decision you make as a solo builder, verified to primary sources.