⚡ Quick Summary
Published by the Centre for Emerging Technology and Security (CETaS) at The Alan Turing Institute, the AI Information Threats and Crisis Response: Practitioners’ Handbook outlines operational actions across the UK government, technology industry, and civil society to counter AI-driven misinformation, deepfakes, and chatbot inaccuracies during acute security incidents. Authors Broderick McDonald and Sam Stockwell structure actionable recommendations across three operational phases—prior to, during, and after crisis events—accompanied by five ongoing policy priorities. The guidance emphasizes structured coordination mechanisms, including cross-Whitehall red-teaming tabletop exercises, dedicated industry crisis command centres, conversational interface fact-checking warnings, model safety patching repositories via the Frontier Model Forum, and community-led rapid response interventions designed to prevent digital threats from escalating into real-world violence.
🧩 What's Covered
The handbook establishes operational interventions structured across chronological phases and institutional sectors:
- Prior to Crisis Events:
- Government: Cabinet Office scenario planning via cross-Whitehall tabletop exercises; updating the Government Communication Service Emergency Planning Framework (PRIMER) to disseminate facts through non-governmental channels (local media, religious centres, sports clubs); and establishing AI Security Institute (AISI) threat intelligence channels and DSIT/NSOIT liaison points.
- Industry: Formalising dedicated crisis command centres and liaison officers; updating Terms of Service under the UK Online Safety Act (Categories 1 and 2B) to downrank deceptive crisis content; implementing prominent user warnings in chatbot conversational interfaces; and expanding Frontier Model Forum (FMF) threat reporting.
- Civil Society: Designing media literacy and digital hygiene initiatives to reduce susceptibility to manipulation tactics.
- During Crisis Events:
- Government: NSOIT threat monitoring focusing strictly on content inciting real-world violence with high evidentiary thresholds; AISI technical advice to the Cabinet Office; and Situation Centre (SitCen) data corroboration.
- Industry: Rapid cross-industry hash-sharing, vulnerability reporting, downranking, demonetisation, and labelling interventions.
- Civil Society: Utilizing automated rapid-response fact-checking tools (SMART 2.0, WITNESS Deepfake Rapid Response Force), platforming credible local community voices, and piloting counter-conspiracy chatbots like DebunkBot.
- After Crisis Events: Formal post-incident reviews modelled after public health response plans, continuous NSOIT indicator evaluations, shared FMF incident databases to guide model safety patching, improved data access for UK researchers, and reporting foreign interference domains to NewsGuard’s FAILSafe platform.
- Ongoing Policy Priorities: Establishing a four-pillar AI preparedness framework (anticipate, prevent, prepare, respond); providing open safety tools (ROOST) and mentorship to small AI startups; issuing user chatbot guidance under DSIT's Digital Inclusion Action Plan; defining severity monitoring indicators for the Commonly Recognised Information Picture; and institutionalising academic engagement.
💡 Why it matters?
During acute crises like terrorist incidents or civil unrest, generative AI tools and conversational chatbots can rapidly amplify falsehoods and incite physical harm. This handbook bridges strategic national security policy and frontline operations, offering practical protocols for trust and safety teams, emergency planners, and civil society actors. By delineating clear communication conduits between frontier AI developers, regulators, and central government, the handbook provides a balanced roadmap that mitigates informational harms while safeguarding fundamental rights and freedom of expression.
❓ What's Missing
The handbook focuses on high-level operational procedures and does not detail technical benchmarking criteria or error-rate thresholds for automated fact-checking and deepfake detection tools. It also leaves open the specific legal mechanisms required to match EU Digital Services Act data access provisions for UK civil society researchers, as well as concrete funding models to resource small AI startups lacking dedicated trust and safety infrastructure.
👥 Best For
AI governance and trust & safety professionals, government crisis communication leads, national security policymakers, social media integrity teams, and civil society fact-checkers responding to emerging information threats.
📄 Source Details
- Title: AI Information Threats and Crisis Response: Practitioners’ Handbook
- Authors: Broderick McDonald and Sam Stockwell
- Organisation: Centre for Emerging Technology and Security (CETaS), The Alan Turing Institute
- Publication Date: April 2026
- Format: Policy Brief / Practitioners' Handbook
📝 Thanks to
Credit to authors Broderick McDonald (The Alan Turing Institute, University of Oxford, King's College London) and Sam Stockwell (CETaS) for delivering practical operational guidance to strengthen institutional and societal resilience against AI-driven information threats.
