Anthropic is putting frontier Claude models and its own engineers behind operational technology security, and launching OSS Scanner for open-source code.

After a two-year supervision programme, the ICO says ten AI developers have changed or will change their data practices. Agentic AI is its next focus.

Crowe's latest Manufacturing Outlook Report shows AI adoption jumping and defence work drawing interest, while costs and skills gaps still hold firms back.

Norwich Apex wants a 150MW data centre at Keswick, beside Marston Marshes. Thousands oppose it, and South Norfolk's council leader says he will object.

Co-op Legal Services now has an OpenAI model grade probate advisers' calls on over 50 measures. Unions object; the Co-op says AI is a support tool only.

Royal Mail used AI for its 2025 Christmas ad slogan. Paul Vaughan's daughter believes it recreates her late father, and analysis finds close similarity.

Vicon's owner paid £525,000 for London startup Move AI's markerless motion capture assets, as weak markets leave its FY26 forecast below expectations.

OpenAI rates the Russian 'Dark Clark' operation its first Category 5 case, while Iran's 'Bogus Bylines' got articles into outlets and targeted Iran International.

Tech Against Terrorism's expanded test of 134 AI models found only two held up, and models answered self-declared researchers far more often than terrorists.

A Pontypool photographer says about a quarter of his school proofs are screenshotted for AI tools, as experts warn of privacy risks to children's images.

A Turing DARe and University of Birmingham model segments solar panels, antennas and thrusters in sub-terahertz radar images to support space domain awareness.

Anthropic's annual policy refresh consolidates its influence-operation rules, sharpens weapons and surveillance limits, and bans needless cruelty to Claude.