Authorities in Pakistan are examining two separate cases involving teenagers after AI chatbot Claude, developed by Anthropic, reportedly flagged conversations that raised concerns about possible violent intentions and shared the information with Federal Bureau of Investigation, which subsequently notified Pakistani authorities. According to investigators, the two 17 year old college students, one from Bahawalpur and the other from Karachi, live hundreds of kilometres apart and have no known connection with each other. However, investigators believe both were influenced by popular American crime drama series and allegedly discussed fictional crime scenarios with artificial intelligence tools. The Karachi case has been reported publicly for the first time, while investigators say both incidents have highlighted the growing role that AI safety systems and international cooperation can play in identifying online conversations that may require further assessment by law enforcement agencies.
According to investigators, the Karachi student allegedly interacted with Claude between September 1 and September 4 while discussing a plan involving an individual he claimed had verbally insulted his father. Officials said the conversations contained references to sedatives, disposal of a body, and the fictional television character Dexter Morgan from the television series Dexter, whom the teenager reportedly referenced while discussing how such a character might approach a similar situation. Authorities said Claude identified the conversation as violating its safety policies and generated an alert that ultimately reached FBI before being shared with Pakistani authorities. During questioning by NCCIA, the student reportedly denied planning any real world violence and stated that the conversations were part of research for a story he intended to write. His father also denied that his son had engaged in any criminal activity, explaining that the teenager spent most of his free time watching television dramas after attending college, where he is enrolled in an information technology diploma programme. Investigators stated that no prosecutable offence has been established in the case, although they continue to monitor developments because of the nature of the reported conversations.
In the second case, investigators said a teenager from Bahawalpur allegedly developed an interest in chemistry after watching the television series Breaking Bad and established a small laboratory at home. According to officials, the teenager reportedly obtained chemicals from different sources and was alleged to have attempted to poison his father following a family disagreement. Investigators stated that after the attempt reportedly did not produce the expected outcome, the teenager consulted Claude on September 12, asking why methanol mixed with food had not produced the intended effect. Officials said the chatbot declined to provide further assistance, after which the teenager allegedly indicated that he would consult other AI platforms, including Grok and ChatGPT. According to the investigation, Claude was the only chatbot that generated an alert leading to notification of FBI. The teenager later denied planning any harmful act, while his father supported his account. Authorities stated that the court subsequently granted bail after legal proceedings, although investigators remain concerned because of the reported experimentation with laboratory chemicals and continue to view the matter as significant from a public safety perspective.
The cases have renewed discussion about the safeguards built into generative artificial intelligence platforms and their ability to identify conversations that may involve potential harm. Investigators referred to industry research indicating that AI platforms differ in how they respond to requests involving violent scenarios, with some systems refusing to assist while others have been found to provide varying levels of information. Officials said the incidents demonstrate the importance of responsible AI development, effective safety guardrails, and cooperation between technology companies and law enforcement agencies when online interactions raise credible concerns. At the same time, the investigations also illustrate the need for careful assessment of AI generated alerts, as authorities must distinguish between fictional writing, academic curiosity, and situations that may present genuine risks. NCCIA officials said developments in artificial intelligence require agencies to strengthen their preparedness and investigative capabilities while ensuring that each case is evaluated on the available evidence and in accordance with the law.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
