In a September 18, 2026 segment, ABC7 Chicago reported that former Homeland Security officials and federal cybersecurity agencies are warning that terrorists and criminal groups are already using AI to improve cyberattacks, craft attack playbooks and learn how to construct weapons. A related SignalNews Chicago article says US adversaries may use AI tools to streamline different stages of terrorist planning.
This article aggregates reporting from 2 news sources. The TL;DR is AI-generated from original reporting. Race to AGI's analysis provides editorial context on implications for AGI development.
The ABC7 and related coverage show how quickly AI assisted terrorism has moved from academic scenario to talking point for former DHS leaders and federal cyber agencies. The core claim is that extremist and criminal groups are already using AI for better target recon, cyberattack planning and weapons know how, not necessarily to build exotic bioweapons but to streamline familiar attack patterns. That aligns with recent research suggesting AI mostly makes low skill actors more capable rather than enabling science fiction capabilities overnight. ([abc7chicago.com](https://abc7chicago.com/post/ai-warning-federal-officials-warn-growing-threat-artificial-intelligence-assisted-terrorist-attacks/19845317/?utm_source=openai))
For AGI timelines, this cuts in two directions. On one hand, visible security concerns can trigger faster regulatory action on model access, logging and high risk query filtering, especially for agents with tool use. On the other hand, the same agencies warning about AI assisted terrorism will likely push for more powerful defensive models in intelligence, threat detection and cyber operations. That dynamic could justify continued frontier scaling even as risks mount.
The most important shift is psychological and political. When mainstream local news starts running segments about AI planning terrorist attacks, it becomes easier for lawmakers to argue that powerful models should be treated more like controlled dual use technologies than like generic software. That could eventually slow or channel the deployment of the most capable systems, even if it does not stop their core technical development.


