How Terrorists Are Using AI, And How The UN Is Responding | Steven Siqueira (UNOCT)

Steven Siqueira (UN Office of Counter-Terrorism) on preparing institutions for AI, the terrorist misuse the UN is tracking, and why testing must be independent. Opening remarks at a FAR.AI convened discussion at the United Nations during UNGA 81, bringing together AI developers, researchers studying the risks, and the counterterrorism community. The speaker sets out both sides of the ledger. Responsibly governed, AI can strengthen analysis, support prevention and build community resilience. At the same time, terrorist and violent extremist actors are using these tools to produce and amplify propaganda, with consequences for recruitment and radicalization. He cites findings presented the previous day by Tech Against Terrorism, whose updated CTI benchmark tested 160 AI models across roughly 100,000 requests and found that many returned detailed answers to harmful requests, including cases where users openly stated terrorist intent. His conclusion is that independent testing has to be repeated as models change, and that the findings have to lead to safeguards administered independently. On weapons of mass destruction, he notes that access to materials, facilities and operational expertise remains essential, while raising the concern that AI systems able to synthesize scientific information and assist with specialized problem solving could lower barriers to certain forms of technical knowledge. The remarks are anchored in the ninth review of the UN Global Counter-Terrorism Strategy, adopted in July 2026, and set out three priorities member states are raising: preparing institutions to use AI responsibly, ensuring counterterrorism experience informs how models are developed, and preventing the weaponization of AI by terrorists. He is direct about the risks on the state side as well, naming discriminatory profiling, unlawful surveillance and unjustified restrictions on freedom of expression as things that must be addressed before deployment, not after. He closes on partnership: that frontier labs need to explain capabilities and limitations in terms policymakers can use, that practitioner experience needs to reach researchers, and that civil society, women and young people identify risks technical evaluations miss. Chapters 0:00 Opening remarks and thanks to FAR.AI 0:17 Where AI helps, and how terrorists are exploiting it 0:53 Tech Against Terrorism: 160 models tested with 100,000 requests 1:17 Independent testing, accountability and anticipatory counterterrorism 1:53 AI, CBRN materials and the barrier to specialized expertise 2:28 The ninth review of the UN Global Counter-Terrorism Strategy 3:13 Three priorities member states are raising 3:41 Institutional readiness: people, processes and culture 4:30 Responsible use and the human rights risks 5:15 How UNOCT supports member states 6:24 Why cross-sector dialogue matters now 7:16 Civil society, women and young people 9:06 Mainstreaming AI across counterterrorism practice Steven Siqueira (UN Office of Counter-Terrorism) argues that AI safeguards require independent testing repeated as models evolve, and that counterterrorism and human rights experience must inform how these systems are built rather than arrive after deployment. More AI safety research: https://far.ai FAR.AI is a research nonprofit working to ensure the safe development of advanced AI. We host the Alignment Workshop series and publish frontier alignment research.