London

Building the foundational AI infrastructure for the era of agentic warfare.

Builds a synthetic battle arena where autonomous defence systems train against adaptive adversaries drawn from live conflict intelligence, compressing years of development into weeks. Aimed at Europe's sovereign defence AI infrastructure; recently secured dedicated access to the UK's national AI Research Resource. Founded by Tristam Constant, formerly Europe director at Anduril and head of European defence at Applied Intuition, and Junaid Hussain, co-founder of an air defence company and founder of the world-building platform Iconic. Cedric Archambeau serves as Chief AI Officer; team draws from Anduril, Applied Intuition, Palantir, Google and Unity. Chaired by Unity founder David Helgason.

Founded 2026

Statutory audit, off your plate.

Runs statutory audits end to end, from the client's accounting systems to the signed opinion. A proprietary platform connects to ledgers and bank accounts, collects the data and handles the volume testing, while qualified auditors review the findings and sign every engagement. Registered as an audit firm rather than a software supplier to other auditors, it already carries out client audits in Norway. Its founders are Thibault Mallion and Mads Bogen Øye, both former technology investors who now run the firm as co-chief executives, and Andreas Rystad, chief technology officer and a former Palantir engineer.

Virtual worlds where robots and AI agents learn.

Builds world models, AI systems that simulate and predict how physical environments behave. The company remains in stealth but is expected to target use cases in robotics and simulation. Everyone named so far worked on DeepMind's Genie project, which generates interactive 3D worlds from a single text or image prompt. Its founders are Jack Parker-Holder, who led the team behind Genie 2 and Genie 3, and two colleagues from the same world-model team, Matthew McGill and Philip Ball, who both contributed to Genie 3.

Founded 2026

Making first contact with superintelligence.

Building a superlearner that discovers knowledge from its own experience rather than human generated data, using large scale reinforcement learning. Founded by David Silver, architect of AlphaGo, AlphaZero and MuZero, with six co-founders almost entirely out of DeepMind, including Lasse Espeholt, Chris Apps and Junhyuk Oh. Advised by Geordie Rose.

Founded 2025

Recursively self-improving to discover new knowledge.

A lab redesigned from scratch to apply recursive self improvement to an entire research organisation, not just one model. A Public Benefit Corporation. Its first result, Faraday, is a 27B parameter AI Scientist that layers scientific intuition onto coding agents to replicate research papers. Tantum Collins, CEO, was a research and strategy aide to DeepMind's CEO before three years in the US government helping establish the AI Safety Institute. Edward Hughes, Chief Scientist, spent nearly nine years at DeepMind. Louis Kirsch did his PhD under Jürgen Schmidhuber before joining DeepMind. Kaloyan Aleksiev, CTO, came from Reka AI.

Founded 2025

Automated Frontier Model Research

Applying recursive self improvement first to AI research itself, with the stated aim of extending the approach to every scientific discipline. Eight co-founders: Richard Socher, formerly Salesforce's Chief Scientist and concurrently CEO of You. com and founder of a venture fund, Tim Rocktäschel of UCL, and six more from OpenAI, DeepMind, Google Brain and Meta. The team built AI labs at Salesforce and Uber, with two prior unicorns among them.

Founded 2026

Robots you can train on the job.

Builds robots that learn new work from the people doing it. A worker shows the task, corrects mistakes as they happen, and every skill is shared across the whole fleet. New behaviours take as little as ten minutes to set up instead of days, with no specialist programming. Plastics manufacturers and electronics disassembly teams have already taught the robots new stations themselves. Its founders are Jonathan Scholz, who built and led Google DeepMind's Applied Robotics team for seven years, and three colleagues from that team: Oleg Sushkov, Misha Denil and Akhil Raju.

Founded 2025 · Funded

A deeper view on AI risk and opportunities across the investment lifecycle.

An AI-native advisory firm for private equity, running buy- and sell-side AI impact diligence and value creation work directly for funds. Ex-MBB consultants are paired with engineers and powered by a proprietary agentic system, rather than sold as software. Works across London, DACH and the US. Founded by Tobias Haefele, who previously built AI products at QuantumBlack, with Zack Zornitta.

Founded 2025

Perfect knowledge is here.

A governed context layer giving enterprise AI agents a shared, continuously updated memory of how a business runs, stored as versioned files the work itself writes back to, with every answer citing its source. Michael Bajwa was Convergence's first product hire, joining at the very start and taking it to its first million in annual revenue within ten weeks, ahead of its acquisition by Salesforce; he previously shipped radiology algorithms at Harrison AI. Qian Zheng was Convergence's first engineering hire, later its Head of Engineering.

Founded 2026

Intelligence forged on the floor.

A network of operators who embed in European precision parts factories serving defence, aerospace and grid infrastructure, rotating through the network before taking founder or operator seats in factories built from the inside. Software ties the network together, tracking cycle times and routing what one factory learns to every other overnight. Matija Bagic spent seven and a half years at Palantir in automotive and industrial. Ognjen Ivkovic spent eleven years at Palantir, most recently as Head of Energy, before a brief stint at OpenAI. Founded in London, 2026.

Founded 2026