A new research report from the Centre for Emerging Technology and Security (CETaS) has warned that artificial intelligence–driven information threats are increasingly being weaponised during crisis events and now pose systemic risks to public safety, national security and democratic stability.
The 130 page study, Adding Fuel to the Fire: AI Information Threats and Crisis Events by researchers Sam Stockwell, Ardi Janjeva and Broderick McDonald, is based on case study analysis, interviews with 25 cross sector experts and a government style tabletop exercise. It concludes that AI tools were used in at least 15 major global crisis incidents since mid 2024, including the Southport riots, the Bondi Beach terrorist attack, the Israel–Iran conflict and the US intervention in Venezuela.
CETaS says that while traditional forms of mis- and disinformation remain significant, AI is enabling faster, more personalised and more scalable manipulation of crisis related information, often with real world consequences for law enforcement, community cohesion and political decision making.
According to the report, generative AI has been used to create and disseminate fabricated images, videos, songs and audio clips designed to inflame tensions, incite violence or distort public understanding of unfolding incidents.
Notable examples include:
- xenophobic AI generated songs and images circulated during the Southport riots, encouraging retaliatory violence;
- a deepfake audio clip of Sadiq Khan in 2023 that police believed came close to triggering serious public disorder;
- synthetic battlefield footage influencing public perceptions during the India–Pakistan and Israel–Iran crises;
- AI upscaled ‘enhanced’ images following the 2025 Charlie Kirk shooting that introduced false details and risked misidentifying innocent individuals.
The report warns that such synthetic content can confuse emergency responders, distort risk assessments and make it harder for the public to access reliable information during fast moving incidents.
Threat actors exploiting AI chatbots and platforms
CETaS highlights a series of cases where AI chatbots embedded into social media ecosystems – including Grok, ChatGPT and Perplexity – incorrectly authenticated AI-generated content, reinforced conspiracy theories or misidentified footage from unrelated events.
Because chatbot interactions are private and not subject to public scrutiny, the report argues they may become a “blind spot” in crisis driven misinformation, with extremist groups exploiting early erroneous outputs to validate misleading narratives.
The research also identifies growing experimentation with data poisoning, with Kremlin linked networks seeking to pollute training datasets to influence future model output, posing risks for automated search summaries and fact checking tools during high stakes incidents.
Despite the scale of the risks, the report finds “worrying uncertainty” around how the UK Government would respond to a serious crisis in which AI generated information was a central threat vector.
Interviewees cited unclear departmental responsibilities, limited mechanisms for rapid information sharing with AI companies, and crisis protocols that are not designed for AI accelerated event timelines.
For local authorities, law enforcement and central government, CETaS argues that the challenge now is not merely technological, but institutional, requiring new protocols, cross sector coordination, and a deeper understanding of how AI mediated information flows can influence public behaviour during moments of acute uncertainty.
The authors recommend:
- clear crisis response protocols led jointly by the Department for Science, Innovation and Technology (DSIT), the Cabinet Office and the AI Security Institute;
- an updated RESIST 2 counter disinformation toolkit incorporating AI specific threats;
- new school and university guidance on the use of AI chatbots for fact checking;
- an Ofcom led evidence gathering exercise on the monetisation of AI driven disinformation, to feed into Online Safety Act implementation.
CETaS also calls on tech companies to establish formal crisis command centres, participate in cross industry incident reporting mechanisms, and introduce clearer pop up warnings on chatbots queried during live incidents.
The report concludes that AI has already reshaped the dynamics of public disorder, geopolitical crises and community tensions, and that without urgent reform “future crisis events are almost certain to be exacerbated” by AI driven information threats.
