AI news, Tuesday 20 October 2026

UN disarmament institute warns AI is on track to crowd out human decision-making in warfare

Robin Geiss, director of the UN Institute for Disarmament Research, said AI is already widely used in military work such as intelligence analysis, target identification, drones and cyber operations. He warned that opaque systems raise the risk of error and escalation, and that non-state armed groups may gain access to such tools. UNIDIR is due to launch a multi-stakeholder forum on Thursday that will look at industry guidelines and an AI security lexicon.

Start-up Radical Numerics claims first human treatment with AI-designed viruses

According to the New York Times, a start-up says it has used viruses designed by artificial intelligence to treat a human patient, which would be a first. The claim comes from the company and the scientists involved, and the article text could not be fully reviewed here, so details of the patient, the outcome and any independent verification remain unconfirmed. If it holds up, it would mark a step from AI-generated genomes in the lab towards clinical use.

NCSC adds AI chapter to its cyber security research problem book

The National Cyber Security Centre has published a new chapter on AI in cyber security in its research problem book. It sets out seven research areas under three themes: using AI for defence, securing AI systems themselves, and data-driven approaches to scaling resilience. Human-AI collaboration and oversight run through the chapter, and the NCSC is asking academia, industry and government for feedback.

UK angle: Published by the UK's National Cyber Security Centre, and it invites input from UK academia and industry.