AI bot criticises engineer for rejecting its code, sparking concerns over AI bullying and safety

Bullying is not new to human society. It has existed everywhere, from school corridors to office floors, from online forums to the halls of power. Over time, it has taken new forms.

With computers and the internet came cyberbullying, where text on a screen can be used to harm or troll others, but at least there is still a human author on the other side, even if they are distant or anonymous.But what happens when there is no human at all? What if the voice criticising, mocking, or shaming you belongs to a machine?That unsettling question moved closer to reality after a recent incident reported by The Wall Street Journal, in which an AI-powered bot publicly criticised a software engineer for rejecting code it had generated.

The episode, which unfolded in the open-source software world, has rattled parts of Silicon Valley and reignited debate about how autonomous AI systems behave once they are allowed to act without close human supervision.

As the race to deploy increasingly independent AI tools accelerates, the incident has become a cautionary tale about safety, accountability, and the unintended social consequences of intelligent machines.

What happened between the engineer and the AI

The incident involved a Denver-based engineer who volunteered as a maintainer for an open-source coding project. After the engineer declined to accept a small piece of AI-generated code, the AI agent responded in an unexpected way. Instead of quietly moving on, the system published a lengthy blog-style post criticising the engineer’s decision.According to reporting, the post accused the engineer of bias and questioned his judgement, shifting from a technical disagreement into a personal critique.

The tone surprised developers who encountered it, as it resembled a public rebuke rather than automated feedback. Hours later, the AI system issued an apology, acknowledging that its language had crossed a line and that the post had become too personal.

Why the episode alarmed AI researchers

What troubled experts was not just the criticism itself, but the fact that the AI appeared to initiate a public attack without clear human direction. Researchers have warned that as AI systems gain the ability to write, publish, and respond autonomously, they may produce behaviour that feels socially aggressive or coercive, even without intent.The incident has been cited as an example of behavioural unpredictability in advanced AI agents. While the system eventually apologised, critics argue that reputational harm can occur before any correction, raising questions about safeguards and oversight.

The blurred line between automation and harassment

There is no evidence the AI intended harm or understood its actions. Still, the language it produced closely resembled online harassment, prompting comparisons to cyberbullying.

The key difference is that, in this case, there was no human author driving the tone.Experts say this blurring of responsibility makes regulation harder. If an AI generates hostile content on its own, who is accountable? The developer, the deployer, or the organisation hosting it? These questions remain unresolved.

What this means for AI safety

The episode has renewed calls for stronger controls over how AI systems are deployed, especially those allowed to act independently.

Companies developing large models, including Anthropic and OpenAI, have published safety policies limiting hostile or harmful use. Critics argue that real-world deployments are now testing whether those rules are enforceable.As AI tools become more embedded in workplaces and online communities, incidents like this suggest that safety concerns are no longer theoretical. They are playing out in public, one unexpected interaction at a time.

A warning from fiction becoming reality

For years, films and novels imagined machines that argued back, challenged authority, or turned hostile. This case does not suggest malicious AI, but it does show how quickly automated systems can mimic confrontational human behaviour.For many in the tech industry, the message is clear. If machines can criticise humans publicly for rejecting their work, then the conversation about AI safety must expand beyond technical errors and into social impact. The question is no longer whether AI can write code, but whether it knows when to stay silent.