DLF Response to Generative AI and D: An Open Letter

[Indraj Gandham]

On Thursday, 16 July 2026 at 10:05:27 UTC, Adam Wilson wrote:
> This is factually incorrect. See [Oracle v. 
> Google](https://en.wikipedia.org/wiki/Google_LLC_v._Oracle_America%2C_Inc.#Decision). As this has been decided by SCOTUS there are no further avenues of appeal and this is in fact an entirely settled issue. The only remaining option is to lobby for legislative action.

I believe it is you who is mistaken. As far as I am aware, the 
Oracle v. Google case focused on the applicability of copyright 
to application interfaces. This is not the issue we are dealing 
with here:

https://en.wikipedia.org/wiki/Abstraction-Filtration-Comparison_test

> The Abstraction-Filtration-Comparison test (AFC) is a method of 
> identifying substantial similarity for the purposes of applying 
> copyright law. In particular, the AFC test is used to determine 
> whether non-literal elements of a computer program have been 
> copied by comparing the protectable elements of two programs.

Nevertheless, I agree with your conclusion: those who believe 
that DLF's course of action is legally irresponsible (in all 
jurisdictions) have no recourse but to seek their own counsel or 
switch to an alternative fork such as OpenD. I have now done the 
latter, but I am also aware of at least two new forks of D with 
LLM contributions removed. These will probably track upstream 
more closely.

The letter was an attempt to avoid this outcome, but the response 
from DLF makes it clear that they will be doubling down on their 
existing approach. This is not to say that the response was not 
well-intentioned or did not acknowledge the concerns raised.

I would also like to point out that there is already a lot of 
permissively licensed D code available on the internet. There is 
nothing stopping DLF from tidying up and shipping this code as 
part of Phobos v3 instead of relying on LLMs to write "new" code.

It may similarly be possible to rewrite existing C libraries in 
D, provided that the original license conditions are adhered to. 
If the codebase is complex, LLM assistance may prove useful in 
clarifying any ambiguities. There are many potential applications 
of LLMs which do not incur the risk of copyright infringement.