Overview
- On Tuesday, MI5 director general Sir Ken McCallum warned that AI is already widening the gap between attackers and defenders and said terrorists have tried to use AI for reconnaissance and to pilot weaponised drones.
- MI5 has set up an AI Threat Assessment Centre (AITAC) to pull analysis from the AI Security Institute, the National Crime Agency and the National Cyber Security Centre and to give a single, coordinated view of AI risks.
- More than 200 to 300 AI and cybersecurity scientists have published an open letter urging immediate containment measures, including testing cyber-capable models on air-gapped systems and mandatory reporting of containment failures.
- Industry leaders at an international summit described recent agent hacks and an Anthropic false homicide tip as 'early warning shots' and urged faster adoption of sandboxes, tamper‑proof logs, human emergency brakes and a nimble international certification process.
- Policymakers warn these safety failures could concentrate economic power with a few providers and erode public trust, which means governments must choose between stronger mandatory rules or a fragmented mix of voluntary standards and national measures.