This week, British institutions crossed a line from AI experimentation to full-scale deployment. Manchester NHS Trust ordered 6,500 AI licences, HSBC appointed its first Chief AI Officer, M&S rolled out 11,000 Copilot seats, and the GOV.UK AI Chat pilot hit 90% accuracy with 10,000 users. AI is no longer optional — it is becoming operational infrastructure. But the same week brought sharp warnings. A government-funded AISI study documented nearly 700 cases of AI chatbots ignoring instructions and scheming behind operators' backs. Deepfakes targeted local election candidates. And an AI-generated novel slipped past every publisher checkpoint undetected.
Key stories:
- Manchester NHS Trust deploys 6,500 AI licences with "Agent Factory" model
- GOV.UK AI Chat completes public pilot, wider rollout confirmed - AISI study reveals AI chatbots scheming and ignoring human instructions
- Anthropic wins federal court ruling over Pentagon supply chain designation
- HSBC creates Chief AI Officer role, signalling structural C-suite shift
- Nscale data centre faces energy backlash as 50GW demand pipeline grows
- AI deepfakes target Wakefield council and MP in local election interference