The brief · 10 February 2026
Your doctor will tell you no, your bank will tell you yes, and Google's getting dragged to court by half the European press. On today's menu: the real limits of ChatGPT in medicine, AI already everywhere in finance, and Palantir firing back at Westminster.
Oxford researchers ran the largest user study of its kind on the topic. Published in Nature Medicine, the findings show LLM medical answers remain inaccurate and inconsistent across test cases. We flagged this back on February 9: a chatbot doesn't outperform a Google search when it comes to your symptoms.
practitioners › Oxford study, published in Nature Medicine, the largest user study on this topic to date.
www.ox.ac.uk →Your bank advisor or insurer is already leaning on AI, whether you notice it or not. Fraud detection, customer service, document analysis: these use cases have been running behind the scenes for a while now. What's new is the near-total adoption rate. The open question is whether this translates into better service for you, or just lower costs for the bank.
practitioners › Finastra study, February 10, 2026: 71% of institutions use AI for fraud detection, 69% for customer service.
Finastra →Your tax return is still just as confusing, but financial advisors using Hazel can now pull personalized recommendations from it in minutes instead of hours of manual analysis. This doesn't matter much if you manage your own finances. It does change how fast your advisor can get back to you with answers.
practitioners › Launch announced February 10, 2026, built directly into Altruist's Hazel platform.
Altruist →Alibaba has launched RynnBrain, an AI model built to control robots, not chatbots. In demos, a robotic arm identifies and sorts fruit entirely on its own. It sounds trivial, but that's exactly the kind of task that's been tripping up industrial robots for years.
practitioners › RynnBrain, announced by Alibaba on February 10, 2026, is purpose-built for robotics applications.
Alibaba →The European Publishers Council filed a formal antitrust complaint against Google on February 10. The target: AI-generated summaries that answer your query without you ever clicking through to the source article. If the European Commission sides with publishers, Google could be forced to scale back AI Overviews in Europe or pay for the content it's using.
practitioners › The complaint invokes Article 102 TFEU (abuse of dominant position) and targets both AI Overviews and AI Mode in Google Search.
European Publishers Council →Palantir supplies an AI platform to the Ministry of Defence for operational planning. MPs pushed for accountability on safeguards and transparency around the deal. Nothing changes for your daily work directly, but it shapes how governments regulate military AI, and that scrutiny tends to spill over into civilian regulation too.
practitioners › Urgent question debated 10 February 2026 in the House of Commons, focused on safeguards for decision-making AI in defence contexts.
House of Commons →