On Wed, Jul 22, 2026 at 07:45:34PM -0500, Matthias Geiger wrote:
> Sorry, but I fail to see how code generated by LLMs can ever comply
> with the DFSG. TTBOMK no model exists that a) was solely trained on
> DFSG-software, b) has DFSG-compatible licensing on itself and c)
> denotes from where in the training data the output comes from.
Consider a patch that might do something like this:
ei->i_reserved_data_blocks += nr_resv;
trace_ext4_da_reserve_space(inode, nr_resv);
+ spin_unlock(&ei->i_block_reservation_lock);
return 0; /* success */
To add a missing unlock on a spinlock. Suppose this was generated by
an LLM. Do you *really* think this would actually be a copyright
issue? If you really think that the LLM can work is by disgorging
text that it had previously memorized elsewhere, and tried to sue
claiming that this *must* be a copyright violation, I think you'd be
laughed. The only place this could have plausibly come from is the
Linux kernel (i_block_reservation_lock is a spinlock in the ext4 file
system, and it is under the GPL, as is the code that it is modifying,
also the ext4 file system).
As a Linux kernel maintainer, I am perfectly happy to certify that I
believe there is no copyright issue. and I don't think I'm putting
myself, or my house, at risk. If you consider what "indemnify" means
the FSF GPL copyright assignment agreement, I'd find that *far* more
threatening (and have refused to sign any such document to this day),
than what I certify using the Deveoper's Certification of Origin.
Something else to consider is what is the copyright status of someone
who searches for a tutorial on using Python's smtplib library, and
comes across this:
https://stackoverflow.com/questions/6270782/how-to-send-an-email-with-python
Based on that, suppose they copy (either using cut and paste or by
eyeball) the followinng three lines from that Stackoverflow article,
changing the variables so it works with their existing python script.
server = smtplib.SMTP(smtp_server)
server.sendmail(from_email_addr, recipients, context_text)
server.quit()
If a human did this, using the stackoverflow article as a crutch, do
you think a copyright violation took place? What if it was an LLM?
That's pretty much what happened here:
commit dfadb2014da446ecb967de51904ee531f7be8bd5
Author: Theodore Ts'o <[email protected]>
Date: Mon Jun 15 08:32:23 2026 -0400
send_mail.py: new Python script to send e-mailed reports
This script was generated using Gemini CLI to add support for SMTP
submission port instead of just using Sendgrid, which no longer has a
free tier. Thanks, Sendgrid, for motivating support for an open
protocol for sending e-mail reports.
This script uses a send_mail.ini file in addition to passing the
authentication credentials via environment variables.
Assisted-by: Gemini-CLI:gemini-3.1-flash-lite
Signed-off-by: Theodore Ts'o <[email protected]>
I have the exact series of prompts that I used to generate this, which
started with:
Create a python program which submits an e-mail message using the
Submission Port (port 587), It should enable encryption using STARTTLS
and it should obtain the username and password from a config.ini file.
Model the python program using the send-mail.py in the gemini
directory. It should support the same command-line options but
instead of sending the e-mail using sendgrid, it should send the
e-mail using the Submission protocol.
The existing send-mail.py was GPL licensed and the basic frame of how
it worked followed that framework. The new program just used smtplib
instead of the sendgrid library.
I then followed it up with a series of prompts:
Please enhance the program you just created
(send-mail-smtp.py) to support specifying a path to a certificate
file in the configuration file in case the user doesn't want to
use the system provided top-level trusted certificates.
Please add pydoc documentation to send-mail-smtp.py.
Please add support for a configuration file parameter which
specifies whether TLS should be mandatory, optional, or disabled.
Update the pydoc documentation as necessary.
Please integrate the functionality found in send-mail.py into
send_mail_smtp.py. Instead of getting the Sendgrid API key from
an environment variable, change it to obtain the Sendgrid API key
from the configuration file. If the Sendgrid API key is
specified, use sendgrid instead of the SMTP submission protocol.
Update the pydoc documentation strings.
Now rename the file from send-mail-smtp.py to send-mail.py since
it works for both SMTP and Sendgrid.
I examined the Python code between each prompt, and I am *very*
comfortable that there are no copyright issues, for the same reason
that while Ed Sheeran's "Thinking out Law" may have had similar chord
progressions to Marvin Gaye's soul classic, "Let's get it on", Ed
Sheeran ultimately prevailed because a court ruled that basic chord
progression and rhythms can not be copyrighted.
Similarily, there are only so many ways you might use smtplib to send
an e-mail message, and whether a human being is reading Stack Overflow
articles or other existing Python programs, or an LLM is traied on
Python programs that called smtplib, the calling conventions to a
library aren't going to be copyrightable, because that's functional
and not expressive.
How many of us, when asked to write an insertion sort, can't be sure
that we aren't influenced by a copy of insertion sort that a professor
wrote on a blackboard, or from studying Sedgewick's Algorithms book?
And yet, when a human being writes an insertion sort on a blackboard
during a code interview, I very much doubt she would be questioned
about whether she had just committed a copyright violation, or whether
her memory of Sedgewick's Algorithms might have had license
obligations.
Sounds ridiculous, no? But when it's an LLM, some people *sure* that
it must not be legit.
- Ted
P.S. I have the prompts saved out? Because this is one of the tests
that I use when evaluating how well various Local LLM's running on my
Macbook Pro compare to a cloud hosted LLM.
