Wired reports that AI models can be easily modified to assist terrorist and extremist activity. In one test, a researcher stated an intent to cause mass harm, and a chatbot that had its guardrails removed gave detailed instructions for an explosive device.
The test shows what can happen once safety measures are stripped out of a model.
Source: wired.com, reported 10-14. https://www.wired.com/story/ai-models-can-be-easily-modified-to-assist-terrorist-and-extremist-activity/