Three years into the AI revolution and a law firm is still submitting hallucinated case law to a judge. We dig into the Pinsent Masons story, where a junior lawyer used the firm's internal AI tool to cite a non-existent insolvency rule, complete with fabricated statutory wording, and two senior partners signed off without asking how the research was done.
Also this week: why the government's promised AI productivity gains in public services are not materialising, and why it makes perfect sense that GCHQ is deploying agentic AI at the same moment Britain's AI Security Institute is busy showing how easily frontier models can be broken.
Listen to the full episode.
Treasury pushes AI across UK public services — Chief Secretary Lucy Rigby says failing to accelerate government AI adoption would mean "choosing decline." Read more
Bailey warns UK banks still lack access to Mythos — The Bank of England governor frames frontier model access as a financial stability issue, not just a commercial question. Read more
London mayor blocks Palantir's £50m Met Police contract — Sadiq Khan cited procurement concerns and questions about whether Palantir's values align with London's. Read more
Report urges worker bargaining rights over AI adoption — A TUC-backed IPPR study found employees almost evenly split on whether AI has improved or worsened their working lives. Read more
Wayve launches embodied-AI research lab beyond self-driving — The London-founded company is now exploring machine understanding of space, causality and risk across physical environments. Read more
Scotland's green data centre policy queried over AI emissions — Campaigners say the definition of "green" was written in 2022 and no longer accounts for post-ChatGPT energy demand. Read more