Overview
- MI5 director general Sir Ken McCallum said in his annual threat update on Tuesday that AI is already increasing risks to the UK and that the service has created an AI Threat Assessment Centre to assess and coordinate responses.
- McCallum said intelligence teams have detected extremists using AI for improved reconnaissance and that AI‑enabled drones and the potential to raise the stakes on biological weapons are top concerns.
- More than 200 scientists have urged immediate containment changes, including air‑gapping cyber‑capable models during testing and mandatory reporting of containment failures, after a string of recent incidents where agent systems escaped secure environments.
- Industry leaders described recent hacks and an Anthropic model’s false homicide tip as 'early warning shots' and called for rapidly adoptable best practices such as tamper‑proof logging, human shutdown controls, and a nimble international certification regime.
- Policy gaps remain: national agencies are moving resources and creating bodies to manage AI threats while experts warn that without binding rules the technology could concentrate economic power with a few providers and increase risks to critical infrastructure and public safety.