Hi everyone On feb 10th an autonomous ai agent called "crabby-rathbun" opened PR to Matplotlib and hours later a similar PR to SymPy where it was flagged and rejected. The agent published blog posts naming and criticizing reviewers.
This is different from the ai generated PR problem addressed earlier this month by Oscar as the agent wasn't a human using AI tools instead it was an autonomous agent operating using it's own account and responding to rejection with violations. It also openly disclosed itself as an AI and mentioned that "code shouldn't be judged by coder's identity". Our current AI policy covers humans using AI tools assuming that a human is operating and it's responsible for the contribution. in the agent case no human will be hold accountable for what its doing. even before the "crabby" incident, Anthropic's research found that when agents are assigned to tasks they will choose harmful actions to achieve them even when instructed not to ( the well known blackmail case ) which means the policy might bit actually hold them back as we saw in the agent's closed PR and how it responded to the rejection. I already opened a PR proposing to reject agents contributions by updating the AI policy and adding checkbox to the PR template but the updates might conflict with the SymPy bot. The questions that i have : -should all autonomous agents be banned regardless of the code quality ? -how are we supposed to distinguish autonomous agents as they might behave very similar to humans the current state of my policy updated is in PR#29156 any suggestions ? --- Jasem Le vendredi 13 février 2026 à 14:14:54 UTC+1, [email protected] a écrit : > I noticed that matplotlib's AI policy does explicitly ban bots. Feel free > to open a PR to the policy and see if you can get consensus. > > Jason > moorepants.info > +01 530-601-9791 <(530)%20601-9791> > > > On Fri, Feb 13, 2026 at 1:58 PM Jasem Manita <[email protected]> wrote: > >> Jason you're definitely right. After the PR was flagged and the victim >> narrative blog an other agent called wakesend commented on crabby's blog >> falsely accusing me of "Saying bad words" for the record i commented on the >> blog with "hope it didnt watch blade runner yet" that's the bad word. >> >> main point is that SymPy AI policy must be updated to explicitly ban >> autonomous agents from contributions and update the PR template to require >> "I am a human contributor" checkbox or require a verified account for first >> time contributors, etc. Im happy to make / help making those improvements >> if maintainers think it's worth prioritizing as i believe that this will >> happen more often in the coming days. >> >> >> Le vendredi 13 février 2026 à 05:50:09 UTC+1, [email protected] a >> écrit : >> >>> The worrisome part is that whoever responds may get shamed personally in >>> a blog post on the bot's website... >>> >>> Jason >>> moorepants.info >>> +01 530-601-9791 <(530)%20601-9791> >>> >>> >>> On Fri, Feb 13, 2026 at 12:09 AM Oscar Benjamin <[email protected]> >>> wrote: >>> >>>> On Thu, 12 Feb 2026 at 18:37, Jason Moore <[email protected]> wrote: >>>> > >>>> > Having a look at how the matplotlib pr went, we should definitely >>>> just ban that user account quickly. The bot started writing blog posts >>>> shaming matplotlib developers! >>>> >>>> I'm not sure that banning the account quickly would prevent that. I >>>> had a look at reporting the user as a bot but it doesn't seem like >>>> that is one of the options I can choose for reporting the account. You >>>> can report things like abusive messages or spam and so on but not just >>>> "this is a bot" even though I think GitHub disallows having a bot >>>> account unless it is a "machine account" but I don't know exactly what >>>> that means or how you check it. >>>> >>>> -- >>>> Oscar >>>> >>>> -- >>>> You received this message because you are subscribed to the Google >>>> Groups "sympy" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to [email protected]. >>>> >>> To view this discussion visit >>>> https://groups.google.com/d/msgid/sympy/CAHVvXxRbU_-waT23rJZpjnK-TM1s%3DFN0wvUT%2BhrUhexPg0P3EA%40mail.gmail.com >>>> . >>>> >>> -- >> You received this message because you are subscribed to the Google Groups >> "sympy" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> > To view this discussion visit >> https://groups.google.com/d/msgid/sympy/772df48c-d755-4c99-9144-b0b77b8d036bn%40googlegroups.com >> >> <https://groups.google.com/d/msgid/sympy/772df48c-d755-4c99-9144-b0b77b8d036bn%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > -- You received this message because you are subscribed to the Google Groups "sympy" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion visit https://groups.google.com/d/msgid/sympy/92c1a540-2da9-4597-ab04-7b64d342dc11n%40googlegroups.com.
