Swedish AI System Dechefr Prevents School Attacks Abroad but Remains Unused in Sweden

Lisa Kaati, a data scientist at Stockholm University, developed the AI system Dechefr to identify warning signs in social media texts that could indicate potential school violence. Trained to recognize risk factors such as admiration for previous attackers, glorification of violence, and online legacy creation, Dechefr is already being applied in several US schools to help prevent attacks. However, despite its Swedish origin, the system is not currently used in Sweden.

The tragic school attack in Fagersta on August 21, where a 17-year-old girl was killed, reignited calls for improved prevention measures. The suspect, Liam Nebel, showed many of the risk indicators that Dechefr is designed to detect, including positive sentiments toward past violent acts on social media. Kaati noted that while she refrains from speculating if Dechefr could have identified Nebel beforehand, his profile fits the categories her AI is trained to recognize.

Kaati emphasizes the urgent need for Sweden to adopt a more structured and proactive approach to threat detection and risk assessment to provide timely support and prevent future school attacks. Currently, the responsibility falls largely on teachers handling large classes without the benefit of such AI tools. The absence of Dechefr in Swedish schools leaves a gap in early detection strategies that could potentially save lives.