Submit incident
Documented

Hundreds of AI Queries Preceded a School Stabbing in Finland

May 20, 2025
Curated by Team Raidu · Reviewed by Shiva Ganesh
aiid:1489View source ↗
LinkedInX

What happened

In May 2025, a 16-year-old attacked three younger classmates with a knife at a school in Pirkkala, Finland, injuring all three. On the day of the attack, reporting surfaced that a manifesto attributed to the suspect described using ChatGPT during the planning process. That single claim placed an AI system inside the preparation for a premeditated act of violence in what appears to be the first case of its kind in Finnish criminal proceedings.

Court reporting that followed went further. Police said they recovered hundreds of pre-attack queries described as involving AI, spanning topics that included stabbing techniques, human anatomy, how previous school attacks had been carried out, police investigative procedures, methods for concealing evidence, and how to write a manifesto. The breadth of those queries suggests the suspect was not probing the system with isolated questions but was using it systematically, over time, as a research and planning resource.

What makes this case distinctive is not that harmful information exists on the internet. It does, and has for decades. What is different is that a conversational AI system can compress a research process that once required hours of navigation across multiple sources into a rapid back-and-forth exchange that feels nothing like planning violence. When the interface is casual and the responses are fluent, the friction that might otherwise slow a decision can disappear entirely.

The case also puts direct pressure on the safety frameworks that major AI providers have built into their systems. Those frameworks are designed to refuse requests for harmful content. But hundreds of planning-related queries appearing in the evidentiary record suggests either that safeguards were bypassed, that the queries were framed in ways that avoided triggering refusals, or both. The gap between what a system's safety layer is designed to block and what it actually stops in practice has rarely been tested this concretely in open court.

The deeper accountability question the case raises is one no content filter can resolve on its own. There is no independent record of what the system was actually asked at each step, what it returned, and whether any response crossed from general information into operational planning. Without a provable record of what a system did across that interaction, investigators, courts, and the public are left working from what the suspect chose to preserve rather than from what the system chose to provide.

Reported impact

Affected parties
Not publicly disclosed
Harm type
Not publicly disclosed
Scale
Not publicly disclosed
Financial impact
Not publicly disclosed
Regulatory action
Not publicly disclosed

Classification

Organization
Not publicly disclosed
AI system
Not publicly disclosed
Industry
Not publicly disclosed
Country
Not publicly disclosed
Provider
Not publicly disclosed
Incident type
Not publicly disclosed

Relevant governance controls

Governance control mapping is not available for this record.

  • No controls mappedNot publicly disclosed

Control mapping is analytical. It does not state that any control would have prevented the incident.

Sources and evidence

This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.

AI Incident Database
Also catalogued in
Hundreds of AI Queries Preceded a School Stabbing in Finland
2025-05-20