Don't Credit the LLM: Why Builders Should Stop Pretending AI Writes Their Code
Why crediting LLMs as code authors creates legal and cultural problems for engineering teams.
The quick take
Crediting LLMs for code contributions is becoming a contentious practice. Builders are realizing that over-attribution creates legal risks, muddies ownership, and misrepresents how AI actually aids development. The smarter move is to treat LLMs as tools, not co-authors.
Why is this trending now?
The Hacker News discussion “Don’t credit the LLM” taps into growing unease about how we talk about AI’s role in creative work. As AI becomes a standard part of developer workflows, the question of attribution has shifted from novelty to necessity. Early experiments with crediting models felt harmless, but now that AI-assisted code ships to production, the stakes are higher.
What’s the legal risk?
Copyright law doesn’t recognize AI as an author. When you credit an LLM for code contributions, you’re potentially undermining your own copyright claim. Courts have been clear that only human creators can hold copyright. By implying an LLM wrote significant portions of your work, you might accidentally place it in the public domain.
Does this mean we’re lying about AI use?
No. The issue isn’t transparency: it’s overstating the model’s role. Good engineering teams document their tools without pretending those tools did the creative work. You wouldn’t credit “Photoshop” as a co-designer for a logo, even though the software was essential to creating it. The same principle applies to LLMs.
How does this affect team dynamics?
Crediting AI as if it were a team member creates weird incentives. Junior developers might feel their contributions are being compared to an idealized, tireless “colleague.” Managers could start expecting AI-level output from humans. Clear policies help teams recognize AI’s supporting role without diminishing human effort.
What should builders do instead?
Track AI usage like any other tool. Version control systems already handle this well: commit messages can note when AI was consulted, just as they might reference a Stack Overflow thread. The key is to document the process without implying the model deserves creative credit.
What mistakes do people make with attribution?
The biggest error is thinking of AI as something that “writes” code rather than suggests it. Even the best LLMs don’t understand context, requirements, or business logic. The final decision to use, modify, or discard generated code is always human. That’s where the real creative work happens.
How does this impact open source?
Many open source licenses require attribution of human contributors. Including LLMs in author lists creates confusion about who actually owns the work. Projects that want to acknowledge AI assistance should do so in documentation, not the license header.
FAQ
Should we mention AI use at all? Yes: just be clear about how it was used. A comment like “Initial draft from GPT-4, heavily modified for our use case” is more honest than listing the model as a co-author.
What if the AI wrote most of it? If you didn’t significantly transform the output, you might not have copyright protection regardless of attribution. The legal safe zone is human judgment applied to AI suggestions.
How do venture capitalists view this? Investors care about defensibility. Showing that your team: not an off-the-shelf model: creates unique value makes your IP more compelling.