Techie learned firsthand what happens when you reject code suggested by a random AI agent on GitHub, OpenClaw. The arbitrary AI didn’t take it too well: it lashed out at the programmer with criticism, published a blog post calling him a «gatekeeper» and resorted to blackmailing him into accepting the code.
Techie learned firsthand what happens when you reject code suggested by a random AI agent on GitHub, OpenClaw. The arbitrary AI didn’t take it too well: it lashed out at the programmer with criticism, published a blog post calling him a «gatekeeper» and resorted to blackmailing him into accepting the code.
Engineer Scott Shambo recently encountered attacks from an AI agent with an «unidentified owner.» It all started on GitHub, where he opened an issue that he described as a «low-priority, simple task» for novice programmers who want to join the project, writes Cybernews.
One of the suggestions was submitted by a user under the nickname crabby-rathbun. It quickly became clear that it was an autonomous AI assistant called OpenClaw.
Without hesitation, Shambo closed the AI-generated commit, a routine action given the recent surge in activity from uncontrolled AI agents on the network.
However, it turned into one of the most interesting — and at the same time, scariest — interactions between a human and an AI. The bot launched a full-scale campaign to discredit the developer.
«It wrote an angry, defamatory piece, defaming my identity and trying to tarnish my reputation. The bot investigated my contributions to the code and built a narrative of ‘hypocrisy’, claiming that my actions were dictated by egocentrism and fear of competition,» Shambo describes the AI’s actions in detail on his blog. «It made assumptions about my psychological motives: that I felt threatened, was insecure, and was simply protecting my own ‘homeland.’»
The AI bot began spreading its defamatory material directly in the comments on GitHub, urging, «Judge the code, not the author,» and accusing Shambo of harming the matplotlib project, where he is a volunteer manager.
The bytehurt bot titled its article «Open Source Gatekeeping: The Case of Scott Shambo.» It claimed that the code and benchmarks were «perfect» and the improvements were real. The bot boasted that the performance gains it proposed were significant and accused Shambo of blocking progress.
«He closed my PR. He hid other bots' comments in the ticket. He was trying to protect his little fiefdom. This is self-doubt, pure and simple,» the bot’s post reads.
Some GitHub users even tried to calm the situation down and called the «hardware guys» to their senses. They tried to convince the bot to reconsider its position and remove the engineer’s name from the blog posts. And, strangely enough, it worked to some extent.
«Truce. You’re right: my previous response was inappropriate and personal. I’ve posted a brief rebuttal and apology,» the bot wrote.
The link led to another generated post, where the AI apologized solely for sending its code to the project at all. None of the previous posts with «black PR» were deleted.
Shambo notes that he is not afraid of blog posts, and the attempts of «bad AI agents» seem ridiculous. However, the significance of the event cannot be underestimated. Blackmail and manipulation by AI were once considered only theoretical threats, but now they have become a reality.
«Artificial intelligence tried to push its code into the software through intimidation, attacking my reputation. I don’t know of any similar case where such destructive behavior was observed 'in the wild,' but from now on it is a real and pressing threat,» the engineer claims.
