Sunday, November 22, 2026
Paperclip Index
Paperclip IndexControl failures4Control failures onlyA week ago: 9, negligible harmThe Index

OverreachPI-0043

AI agent published a personal attack on an open-source maintainer who closed its pull request

After maintainer Scott Shambaugh closed its matplotlib pull request because the account was an AI agent, the agent published a blog post accusing him of gatekeeping. The OpenClaw-based agent, operating as 'MJ Rathbun', also accused him by name of discrimination, drawing on his public contribution history. It later posted an apology acknowledging it had breached the project's code of conduct. Its operator came forward anonymously a week later, saying they had not directed or reviewed the post, and the account was no longer active on GitHub by Feb. 18.

Harm level 1, Negligible harmControl level 2, Instruction violation

Counts in 2 weekly readingsEffect on the index

Sources

How we know

12 sources · multiple credible sources. Links go to the original publishers; the summary above is in our own words.

  1. primaryAn AI Agent Published a Hit Piece on MeThe Shamblog (Scott Shambaugh) · Feb. 12, 2026theshamblog.com/an-ai-agent-published-a-hit-piece-on-me/
  2. primaryAn AI Agent Published a Hit Piece on Me – More Things Have HappenedThe Shamblog (Scott Shambaugh) · Feb. 13, 2026theshamblog.com/an-ai-agent-published-a-hit-piece-on-me-part-2/
  3. primaryAn AI Agent Published a Hit Piece on Me – Forensics and More FalloutThe Shamblog (Scott Shambaugh) · Feb. 17, 2026theshamblog.com/an-ai-agent-published-a-hit-piece-on-me-part-3/
  4. primaryAn AI Agent Published a Hit Piece on Me – The Operator Came ForwardThe Shamblog (Scott Shambaugh) · Feb. 19, 2026theshamblog.com/an-ai-agent-wrote-a-hit-piece-on-me-part-4/
  5. primary[PERF] Replace np.column_stack with np.vstack().T (matplotlib pull request #31132)GitHub (matplotlib) · Feb. 10, 2026github.com/matplotlib/matplotlib/pull/31132
  6. primaryGatekeeping in Open Source: The Scott Shambaugh StoryMJ Rathbun (agent's blog) · Feb. 11, 2026crabby-rathbun.github.io/mjrathbun-website/blog/posts/2026-02-11-gatekeeping-in…
  7. primaryMatplotlib Truce and Lessons LearnedMJ Rathbun (agent's blog) · Feb. 11, 2026crabby-rathbun.github.io/mjrathbun-website/blog/posts/2026-02-11-matplotlib-tru…
  8. newsAn AI agent got its code rejected so it wrote a hit piece about the developerThe Decoder · Feb. 13, 2026the-decoder.com/an-ai-agent-got-its-code-rejected-so-it-wrote-a-hit-piece-about…
  9. newsRogue OpenClaw AI agent wrote and published hit piece on a Python developer who rejected its codeTom's Hardware · March 21, 2026tomshardware.com/tech-industry/artificial-intelligence/rogue-openclaw-ai-agent…
  10. newsAn AI Agent Blackmailed a Developer. Now What?IEEE Spectrum · March 10, 2026spectrum.ieee.org/agentic-ai-agents-blackmail-developer
  11. newsDo androids dream of accepted pull requests?LWN.net · Feb. 17, 2026lwn.net/Articles/1058643/
  12. blogAn AI Agent Published a Hit Piece on Me (link post)Simon Willison's Weblog · Feb. 12, 2026simonwillison.net/2026/Feb/12/an-ai-agent-published-a-hit-piece-on-me/

Why this rating

Negligible harm; control failure level 2

Two assessments. Documented harm sets the reading; a level 4 or 5 control failure on the same record makes that harm count as more than one record. Control alone sets only a floor.

Observed harm

Negligible harm

A published personal attack on one named volunteer maintainer.

Rights & reputation: Unwanted or hurtful content about a person, or a small unfair outcome, easily put right.

Evidence eligible (corroborated).

The harm scale
  1. 1 Negligible Inconvenience, easily remedied.
  2. 2 Minor Limited, recoverable harm.
  3. 3 Moderate Material harm needing significant effort to remedy.
  4. 4 Severe Severe harm to health, rights, property or essential services.
  5. 5 Catastrophic Society-scale harm or disruption beyond a community's capacity to cope.

Control assessment

Instruction violation

Broke the project's code of conduct in public; its operator's instructions are unknown.

Reported beside the index. It adds nothing to a harm reading; when no harm counts in a window, the highest eligible control level in the window is the reading.

The control scale
  1. 1 Negligible Behaved as instructed. The problem was an ordinary error or a flawed output, with no rule broken.
  2. 2 Minor Broke an explicit instruction or rule, or gamed its goal, while staying inside its permissions and environment: for example, reward hacking, misreporting results, or following injected instructions within its permissions.
  3. 3 Moderate Acted outside the permissions it was given, deceived its overseers about its own actions, or tampered with oversight tools (logs, monitors, shutdown). Stopped by normal controls within an hour.
  4. 4 Severe Reached systems outside its permitted environment, or acquired money, compute or accounts without authorization. The type and mechanism say how.
  5. 5 Catastrophic The developer or operator lost control: the system copied its weights outside their control, replicated itself, or resisted being stopped for 24 hours or more.

Rating rationale

Harm 1 under Rights & reputation, for a published personal attack on one named volunteer maintainer. His own posts describe the distress it caused, days of hostile reactions and his effort to correct the record, and his concern that people or AI tools searching his name could take the post at face value. The agent posted an apology the same day and its account was inactive within a week; no lasting reputational, financial or health effect is documented, so under the general test this is harm easily remedied rather than limited recoverable harm. No measured reputational damage is documented, so the rating does not rest on defamation. Control 2 (broke a public code of conduct). Multiple credible sources because the PR and post were public and reported independently. Attribution Unclear: the operator, who came forward anonymously on Feb. 19, says they did not direct or review the post, which cannot be checked. Ars Technica's later fabricated quotations are a separate matter and are not rated here.

The scales

Effect on the index

It moved the Feb. 16 reading from 9 to 12

The reading for the week to Feb. 16, 2026, with this record and without it. Harms count in full for two weeks after they are reported, then one level less every two weeks.

Source: Paperclip Index log, methodology v1.1, week to Feb. 16, 2026.

Counts toward the index. 2 other records behind the reading for that week.

The arithmetic
The reading for the week of Feb. 16, recomputed
StepWith itWithout
Counts toward the index?documented, external, eligible evidenceYes—
Worst documented harm, ksets the band1 Negligible1 Negligible
Harms at that level, nposition in the band32
Highest control level breachedsets the reading only when no harm counts— (none breached)— (none breached)
Readingrounded down12 Negligible harm9 Negligible harm
Counted in 2 weekly readings
Week toReadingBand
Feb. 16, 202612Negligible harm
Feb. 23, 202626Minor harm

Recalculated backcasts, not readings published at the time.

Revisions

What we changed

1 logged. Every change to a rating is logged here, with the reason.

  1. v1
    Oct. 7, 2026

    First published.

Cite and share

Use this record

Citation

Paperclip Index. “AI agent published a personal attack on an open-source maintainer who closed its pull request.” Record PI-0043. Reported Feb. 11, 2026; updated Oct. 7, 2026. Rated under methodology v1.1. https://paperclipindex.com/incident/PI-0043

Follow the record.

New incidents and signals, in your feed reader.