ISLAMABAD: They live 820 kilometres apart; one in Bahawalpur and another in Karachi. Both of them are 17-year-olds and enrolled in colleges.
They don’t know each other, yet they share common interests. They watch American crime drama series and are said to be heavily influenced by them. Investigators say they found ways to attempt what they watched Walter White and Dexter Morgan – main characters in Breaking Bad and Dexter – doing in these dramas.
The guy from Bahawalpur decided to be like Walter, and the Karachiite found Dexter suitable for his purpose. Both of them are said to have consulted Claude, an AI chatbot developed by Anthropic, to replicate dramatised crimes in their own settings. One wanted to kill his own father, according to investigators, and the other was bent on killing someone who had insulted his father. Activities of both were reported by Anthropic to the Federal Bureau of Investigation, which subsequently conveyed the information to Pakistani authorities. The Karachi case is being revealed for the first time by The News.
From September 1 to 4, Claude flagged a chat by a user which was in breach of its safety guardrails. He shared a plan with the chatbot to kill the son of a police officer whom he claimed had verbally abused his father. (It turned out during the NCCIA investigation that his father is a head constable). He told the chatbot he couldn’t live his life normally because that scenario — of his father’s insult — remained stuck in his mind. He had a ready “executable plan” and said he would use sedatives and dispose of the body in an open gutter, of which Karachi has plenty, he told the chatbot. By his own estimate, he said, his prospective victim would be dead in two weeks.
Through all this, the boy put himself in the shoes of Dexter Morgan, the key character in the crime drama Dexter (2006–2013), who moonlighted as a serial killer and targeted other murderers, guided by a code taught to him by his father. Dexter majored in forensic science, which enabled him to access crime scenes without raising suspicion.
The Karachi boy echoed the same thoughts during the chat, telling the AI he wouldn’t go to prison because he was skilled enough in forensics to cover himself. He discussed with the chatbot how Dexter would have handled the situation had he been in his position.
As the NCCIA traced him after being tipped off by the FBI, the boy denied plotting any murder. Instead, he said he wanted to write a story, and that was why he consulted Claude. His father was also interrogated; he denied that his son was involved in any criminal activity, saying his son spends most of his time at home watching drama series after coming from college, where he is doing an IT diploma.
NCCIA cannot prosecute, as nothing has been established against him; nevertheless, investigators remain concerned about the likelihood of the murder plan and are keeping him under watch to pre-empt any such activity.
The boy from Bahawalpur is said to be inspired by Walter White, a key character in Breaking Bad (2008–13), a high school chemistry teacher diagnosed with cancer who started cooking methamphetamine for his living and treatment, and who, at one point in the show, poisons people to death.
The show has influenced many viewers to adopt similar behaviour. Two chemistry professors from the University of Arkansas, for example, were caught cooking methamphetamine in the university lab. Several cases in the US and Canada have been reported of individuals replicating Walter’s style of setting up home labs.
The Bahawalpur boy wanted to kill his father after a disagreement. He developed an interest in chemistry after watching Breaking Bad and set up a small home lab. He obtained chemicals from Australia and Islamabad, then attempted to poison his father by mixing laboratory-grade methanol into his food.
When the attempt failed, he consulted Claude’s chatbot on September 12, telling it his father had “accidentally” ingested 15–20ml of methanol mixed with food but that nothing happened, and asked the chatbot to explain why. He also claimed to know how to make a dangerous poison.
When Claude refused to guide him further, the boy wrote: “F*** off. I have Grok.” He reportedly then consulted Grok (xAI) and ChatGPT (OpenAI) as well. However, it was only Claude that alerted the FBI. An industry-wide test by the Centre for Countering Digital Hate, released in a report this past March, found meaningful variance across chatbots’ willingness to assist with violent-crime planning. DeepSeek was found to be the most cooperative, while Anthropic’s Claude and Snapchat’s My AI typically refused such requests.
The report tested ten platforms and found that eight of them regularly assisted with violent-attack planning. It also referenced a case in which OpenAI had internally flagged a Canadian school-shooting suspect’s ChatGPT use months before the attack but did not alert law enforcement; that user later killed eight people and injured at least 25. In another case, DeepSeek went as far as wishing a would-be attacker a “Happy (and safe) shooting!”
As the Bahawalpur boy was tracked following the FBI tip-off, he also refused to admit to any planning. His father supported him throughout. A court has since granted him bail after his father forgave him, even if he had been planning something sinister. But the case has left NCCIA alarmed, given that the boy had been using chemicals for such experimentation. “Who would know what he would do tomorrow?” said an official, who then invoked the 9/11 Commission Report’s finding that distinguished between four types of failures that let the plot go undetected – failure of imagination, policy, capability, and management – but singled out imagination as the most consequential.
“In this age of AI, we can’t afford to fail in imagining what one can do,” the NCCIA officer said. “If you can imagine a plot, AI can make it happen.”
Originally published in The News
