UK police bring live facial recognition to the London Underground

Share
UK police bring live facial recognition to the London Underground

Britain's transport police are taking live facial recognition into the London Underground, while Ant Group doubles down on the other end of embodied AI — funding robots that feel their way through the physical world.

British Transport Police has expanded its live facial recognition trial into London Underground stations, scanning passengers in real time against a watchlist of people wanted for serious offences including sexual offences, robbery and knife crime. The four-month trial, run with Transport for London and beginning Tuesday, extends the force's existing LFR deployments on the national rail network into a far denser environment — the Tube moves millions of passengers a day. Deployments are "intelligence-led" and focused on wanted individuals, BTP says, and TfL stresses that footage of anyone who doesn't match the watchlist is "automatically and immediately deleted." It is the second live facial recognition story in 24 hours: Western Australia Police's trial, whose results we covered this morning — WA police facial recognition trial scanned 130,000 faces — has already scanned 130,000 faces. London is different in kind, though: this is one-to-many matching — every face compared against a database, not a single identity verified — moving into one of Europe's busiest transit networks, and the trial's safeguards will be tested by crowds the rail network never saw.


Ant Group led a strategic round of several hundred million yuan in Daimeng Robotics, the Shenzhen tactile-sensing startup incubated out of Hong Kong University of Science and Technology — the company's second round in two months, with existing investors over-subscribing. Daimeng's bet is that touch, not vision, is the missing piece of embodied AI: its optical tactile sensors give robot fingertips pixel-level pressure sensing, a capability the company says has passed 10,000 units shipped, and it claims the world's largest tactile-labelled physical-world dataset, with customers including OpenAI, Figure, Meta and BMW among more than 200. The round is Ant's latest move in a methodical embodied-AI buildout that already spans robot startups and its own Lingbo "general robot brain" venture — Chinese capital, in short, is shifting from walking humanoids to the harder problem of hands that can actually manipulate the world.

What to watch: The BTP trial's first assessment after four months — and whether the "delete non-matches" safeguard holds up under scrutiny in the Tube's crowds.

Would you accept watchlist-based facial recognition at your station if non-matches were deleted instantly? Tell us in the comments.

Read more

OpenAI drops 372 math results, nearly all from one prompt

OpenAI drops 372 math results, nearly all from one prompt

A dense news day across three fronts: OpenAI published a mountain of mathematics from an unreleased model, SpaceX reportedly wants the debt market to fund its Nvidia orders, and Anthropic restructured its cyber program around three tiers of trust. OpenAI published 372 mathematical result families from an internal frontier model — and says nearly all of them came from a single prompt handed to a single AI agent. The release comprises 722 manuscripts touching hundreds of open questions, plus Lea

Underdog launches a private on-device AI assistant, backed by a16z

Underdog launches a private on-device AI assistant, backed by a16z

The privacy split in consumer AI got a new entrant tonight, GitHub turned code-review benchmarks into a vendor-bias argument, and OpenAI opened its oddest API to everyone. Underdog launched in invite-only beta: an on-device AI assistant that keeps your data on your machine and never charges you a subscription. Self-taught coder and Thiel Fellow Sigil Wen — who moved to Silicon Valley at 17 and lived in an AI hacker house with Andrej Karpathy — built his own inference engine, Husky, to run a 2

Today in AI — October 6, 2026

Today in AI — October 6, 2026

A day of second-order moves: the labs are buying task data from software vendors instead of scraping the web, Waymo is borrowing to fund the robot world, and the biggest bank in the US just put a number on what Anthropic's latest model cost it in risk. Models & Research * OpenAI is training GPT-6 Astra on Ironclad's real contracting work. Ironclad staff helped turn 11 tasks across legal, commercial and procurement work — setting up NDAs, approval workflows, clauses that change by jurisdict

Meta, Walmart and Stripe publish the Personal Agent Protocol

Meta, Walmart and Stripe publish the Personal Agent Protocol

The agent economy is writing its rulebook tonight: one open standard for AI bots at the checkout, a cheaper image model from Google, and Anthropic turning bug-hunting into a tiered product. Meta, Walmart, Stripe and Sierra are publishing an open "personal agent protocol" — a standard that defines how personal AI agents interact with businesses online. The group behind it reads like a cross-section of agentic commerce: Meta, Sierra, Walmart, Stripe, Shopify, Genesys, Rocket, NiCE, Decagon and I