EFF policy says bots can code but humans must write the docs
- Reference: 1771592156
- News link: https://www.theregister.co.uk/2026/02/20/eff_demands_human_documentation_as/
- Source link:
Software engineer reveals the dirty little secret about AI coding assistants: They don't save much time [1]READ MORE
The venerable non-profit's Alexis Hancock and Samantha Baldwin [2]laid out its policy this week , explaining: "We strive to produce high quality software tools, rather than simply generating more lines of code in less time."
And while "LLMs excel at producing code that looks mostly human generated" they pair said these models often have "underlying bugs that can be replicated at scale."
"This makes LLM-generated code exhausting to review, especially with smaller, less resourced teams."
Or put another way, "well intentioned people" submit code with hallucinations, submissions, exaggeration or misrepresentation.
[3]
This means project maintainers can end up refactoring, not just reviewing code, when a contributor doesn't actually understand the code they’ve generated. And that’s when the code is actually useful.
[4]
[5]
The EFF is clearly worried about a tidal wave of AI generated code that is "only marginally useful or potentially unreviewable."
So, it is asking contributors to "please refrain from submissions that you haven't thoroughly understood, reviewed, and tested." Baldwin and Hancock want them to disclose when their contributions came courtesy of an LLM and insist that documentation and comments is "human" in origin.
[6]
"Project leads can determine if submissions aren't reasonably reviewable."
OpenUK CEO Amanda Brock said the open source community was only now grasping the impact of AI-generated code on projects and maintainers.
"It's a combination of the breadth of content that's scraped and number of bots doing it, and the volume of contributions and level of 'garbage in' when we look at project contributions."
[7]AI agents can't teach themselves new tricks – only people can
[8]Agile Manifesto turns 25 – just in time for vibe coding to test it
[9]Copilot spills the beans, summarizing emails it's not supposed to read
[10]Your AI-generated password isn't random, it just looks that way
[11]Anthropic tries to hide Claude's AI actions. Devs hate it
She predicted: "We're gonna see a lot more AI red cards in the coming weeks."
The EFF currently has four projects on the boil, all under GNU or Creative Commons terms: Certbot; Privacy Badger; Boulder; and Rayhunter. Needless to say, they all have a privacy and security bent.
[12]
It's no surprise the EFF's policy change has a slightly world-weary tone. Hancock and Baldwin said "It is worth mentioning that these tools raise privacy, censorship, ethical, and climatic concerns for many. These issues are largely a continuation of tech companies' harmful practices that led us to this point.
"We are once again in 'just trust us' territory of Big Tech being obtuse about the power it wields," they added. ®
Get our [13]Tech Resources
[1] https://www.theregister.com/2025/11/14/ai_and_the_software_engineer/
[2] https://www.eff.org/deeplinks/2026/02/effs-policy-llm-assisted-contributions-our-open-source-projects
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aZiTNs7BH6GFd-7mXQZeYAAAANY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aZiTNs7BH6GFd-7mXQZeYAAAANY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aZiTNs7BH6GFd-7mXQZeYAAAANY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aZiTNs7BH6GFd-7mXQZeYAAAANY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.theregister.com/2026/02/19/ai_agents_cant_teach_themselves/
[8] https://www.theregister.com/2026/02/19/jon_kern_vibe_coding/
[9] https://www.theregister.com/2026/02/18/microsoft_copilot_data_loss_prevention/
[10] https://www.theregister.com/2026/02/18/generating_passwords_with_llms/
[11] https://www.theregister.com/2026/02/16/anthropic_claude_ai_edits/
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/front&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aZiTNs7BH6GFd-7mXQZeYAAAANY&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[13] https://whitepapers.theregister.com/
Sic
>will draw the line at non-human generated comments and documentation
And while "LLMs excel at producing code that looks mostly human generated" they pair said these models often have "underlying bugs that can be replicated at scale.
Can we therefore draw a conclusion from this?
Worst of Both Worlds?
If there’s one thing that decades in the industry has taught me, it’s that we software engineers aren’t great at writing useful documentation.
So shoddy code AND terrible documentation?
Of course the best option is clean, readable code written by an experienced engineer. Minimal documentation required - mostly just the public APIs.
Re: Worst of Both Worlds?
Nope ...
There is no excuse or 'magic' process that makes 'Documentation' no longer needed !!!
I really don't understand the problem with documentation !!!???
It is a useful way of understanding what you have produced and going through the logic again while you write the documentation.
It helps to highlight assumptions and errors in the original process AND most usefully it helps the Customer to use the 'thing' you produced.
Helping the customer use the product reduces the amount of 'Customer Support' needed and actually helps to sell more of the product !!!
Word of mouth sells the quality of the product and the quality of the support (because you needed so little of it !!!)
The more you create documentation the easier it gets ... until it is no longer the 'chore' you thought it was !!!
:)
Re: Worst of Both Worlds?
When I began my first software development job (decades ago), the first task allocated to me was writing user documentation for software which hadn't yet been written.
Based on the customer requirements, this meant that we had to understand and clarify the requirements early in the development process. Deficiencies & contradictions in the requirements were identified & corrected. This draft documentation was of course updated as required before delivery to the customer along with the code.
Re: Worst of Both Worlds?
There is indeed something known as "requirements engineering" but I guess it got thrown out with the trash years ago. To be fair, done right "user stories" can help.
Re: Worst of Both Worlds?
" There is no excuse or 'magic' process that makes 'Documentation' no longer needed !!! "
This.
There is very little code around with a suitable level of commenting that no documentation is necessary.
And when you get those people who think that comments aren't necessary because the code is obvious and self documenting? Grrrrr! That's the worst sort of mess to have to delve into, especially if the programmer spent far too much time "doing clever things" so there's the code, the total lack of comments on WTF is going on, and all sorts of possible potential side effects. The last time I came across code like that, I looked at what the API was trying to do and just rewrote all the damn functions. Sometimes life is too short......
Re: Worst of Both Worlds?
The comments (embedded, within the code, not in separate documents or header file) need to explain the PURPOSE of the code - not just what it does, WHY it does it this particular way, and under what conditions it can be executed. Those are concepts that can not be expressed in any programming language I know.
Without that information, the code is un-reviewable and un-maintainable, EVEN LATER BY THE PEOPLE WHO WROTE IT.
Just asking ..... for when there be no comments accompanying and explaining the dislike vote
Does El Reg tolerate and approve non-human generated down votes on comments?
Re: Just asking ..... for when there be no comments accompanying and explaining the dislike vote
You'd have to be pretty sad to create a bot to go downvoting a bunch of comments on a tech website.
Though, maybe this explains "that string of one downvote" on comments critical of a certain drug-using techbro? Yeah, he's about sad enough and has the money to pull it off. Shall I expect to see one lonely downvote just here:
Re: Just asking ..... for when there be no comments accompanying and explaining the dislike vote
Don't tempt them
Documentation, meh. Show me the tests
YMMV but when reviewing software I like to look at the tests to see if the code does what it says it should. If the tests are okay, and assuming they pass, then they can form the basis of documentation.
Re: Documentation, meh. Show me the tests
Upvoted but with a proviso. You need to be sure that the right things are being tested.
Some years ago I made a bug report giving details of how to duplicate it. Specific controls had to be set in a certain way. To their credit, it was fixed fairly quickly but I got a note back from one of the developers :-
"We didn't think anyone would do that."
LLM
> Now create documentation and comments. Make sure they look like human.
> I have generated the documentation and comments now, they look like human as requested.
O
\
/\
O
\
/\