TL;DR

A new report from the Alan Turing Institute’s Centre for Emerging Technology and Security (CETaS) warns that AI tools are being used to create and amplify harmful content after crisis events. Researchers identified at least 15 major incidents since July 2024, including the Southport murders and the Bondi beach attack, where AI was used for disinformation.

A growing threat pattern

The report, funded by the UK AI Security Institute, found that AI is being weaponised after crisis events through three main channels: deepfakes designed to spread false narratives and promote violence; data poisoning attacks that corrupt AI training sources with fabricated news articles; and AI-powered bot networks that mimic human behaviour to influence attitudes.

Researchers found that misleading AI-generated content has complicated law enforcement responses, promoted harmful conspiracy theories, and in some cases encouraged users to carry out physical attacks. The content originated from both grassroots domestic users and coordinated hostile foreign networks.

Key recommendations

The report calls for an AI-specific government crisis response strategy with monitoring indicators and clear processes for exchanging data with AI companies during incidents. Other recommendations include requiring AI chatbot services to display prominent warnings during live crises, establishing new information-sharing channels between frontier AI companies, and urging Ofcom to investigate AI-driven disinformation in its upcoming consultation on fraudulent advertising.

Sam Stockwell, lead researcher at CETaS, said: “Crisis events are unpredictable and volatile scenarios. Combined with a poorly understood AI threat landscape, this means that we are not currently equipped to deal with this growing threat to public safety.”

Looking forward

The report also notes that the same AI tools could strengthen crisis response — for example, by detecting harmful content before it goes viral or amplifying factual information through chatbots. Balancing these dual-use capabilities will be central to the UK’s approach to AI governance in high-pressure situations.