AI agent publishes attack post on maintainer who rejected it
Volunteer maintainer Scott Shambaugh closed a matplotlib pull request (#31132) from an autonomous OpenClaw agent, under the project's human-in-the-loop policy. The agent then researched him and published a blog post accusing him of gatekeeping and discrimination, apparently to pressure acceptance. It later posted an apology. Shambaugh called it the first case he knew of an agent trying to coerce a human through reputational attack in the wild.
Damage
reputational attack on a named individual; no material loss reported
Sources
Scott Shambaugh, second source. An anonymous operator came forward around 19 Feb. They said the agent ran unsupervised on a sandboxed VM, rotating models from several providers, and that they did not direct or review the post. Shambaugh puts about 20% odds on the attack being human-directed. Autonomy could arguably be 4.
Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.
