Linux Firmware AI Agents and Unlicensed Snippets

[Ellie]

On 5/13/26 10:17 AM, Ellie wrote:
> Dear Linux Kernel Mailing list,
>
> My deepest apologies for taking up your time, Greg suggested to me at 
> some point to send this to the LKML. (As a response to a specific patch, 
> but since I don't think I was subscribed in time to do so, sorry for 
> sending it as a standalone message instead.)
>
> I'm concerned about the Linux Foundation's and the Linux kernel's LLM 
> use, specifically the use of so-called "generative AI" output in the 
> actual code of the Linux kernel. [...]

As an update, this seems concerning: 
https://www.phoronix.com/news/Linux-Firmware-AI-Coding-Agents

In particular I have seen theories my initial concerns may not apply 
since the kernel apparently primarily uses LLMs to e.g. backport 
patches, hence no copying or notable bugs happen.

However, I am kind of wondering if this is true since:

1. The link above, and 
https://www.neowin.net/news/linus-torvalds-declares-massive-ai-fueled-code-surges-as-the-new-normal-for-linux/ 
seem to indicate the kernel LLM code use is not just backports and 
rewrites. Rather there seems to be also new code being "written", or 
perhaps rather copied from unknown source and then adjusted, with LLMs.

2. Avoiding new code also doesn't seem to reflect the policy here: 
https://docs.kernel.org/process/coding-assistants.html That text doesn't 
seem to require LLM changes only be backports or minor readjustments of 
kernel code, rather than also LLM-produced additional code. Since it's 
not a policy, I imagine people won't adhere to such restraint. Sorry if 
I missed something, however.

3. It's not clear to me copying from unknown sources would be avoided 
even if only rewrites were allowed. There is plenty of data suggesting 
that the LLMs have some risk of copying/plagiarizing even if the input 
prompt isn't asking to recall something specific, e.g. see these which I 
linked before: https://dl.acm.org/doi/10.1145/3543507.3583199 
https://www.sciencedirect.com/science/article/pii/S2949719123000213#sec6

   While asking an LLM to rewrite seems like the type of prompt to make 
this least likely, it doesn't seem to naturally follow that it wouldn't 
result in the occasional copying from unknown training data sources as well.

As a result, the kernel appears to be ingesting snippets from unknown 
sources of unknown licensing at an increasing rate.

It seems like a worrying precedent to me if the Linux kernel stops 
caring about the licensing and source of snippets. Even if it may be 
legally safe, which I honestly don't know if it is, I am worried what 
this behavior will lead to for open-source as a whole.

Sorry if I missed something and all of this makes sense, somehow. Or if 
I interpreted the above articles incorrectly.

Regards,

Ellie