The question that started it
Erin McGurk and David Khachaturov met at the University of Cambridge — through Caledonian dancing. She was studying Land Economy, having arrived from Altrincham Girls' Grammar School, where she studied from 2017 to 2024. He was completing a doctorate in machine learning. It is not the usual founding story, and that is part of the point: the two of them bonded not over a market opportunity but over a shared concern about what AI assistants were quietly doing with personal information.
The concern was specific. Assistants were collecting, storing and sharing details about people's lives — preferences, schedules, relationships, health, work — and users had almost no say in where that information lived, how it got used, or who it was used by. Every new tool wanted the same life story explained again, and each one kept its own copy. McGurk and Khachaturov decided that was backwards: the person should own the context, and the tools should ask permission.
Two founders, one thesis
The pair's backgrounds explain why they saw the problem the way they did. McGurk, the CEO, came at it from the human side: Land Economy at Cambridge, a degree about how people, land and value interact. Khachaturov, the CTO, came at it from the machine side: a PhD in computer science from Cambridge, where he also held academic posts including at St Catharine's College, then machine-learning research at OpenAI and leading AI security research at FrontierOne. One founder understood what was at stake for the user; the other understood exactly how the systems worked.
Their thesis, as McGurk stated it when the company went public, is disarmingly simple: AI should work in the interests of the person using it. That became Egoist Machines — a Y Combinator-backed startup, still just the two of them, focused on permission-based systems for sharing AI data. The name is deliberate: the ego, the self, is the point. Your data should serve you, not the apps that collect it.
The beta and the $3M seed
Egoist Machines spent 2026 in the open, but quietly. A beta version of AI Passport went out earlier in the year with a waitlist, and the company opened the programme while McGurk was still, in the local paper's words, a former Altrincham student building a new AI company in California. The public launch came on 7 October 2026 — and it came with money: a $3 million seed round with participation from Y Combinator, Goodwater Capital, a European family office, and scientists from DeepMind and Meta.
The timing mattered. October 2026 was the moment AI assistants stopped being novelties and became daily infrastructure — which meant the question of who owns the context inside them stopped being theoretical. Egoist's launch release put it bluntly: AI Passport is “the only way to make personal context portable across AI assistants like ChatGPT, Claude, Muse and Gemini — but still owned by you, not by the apps that collect it.” The “only way” claim is the company's own; what is verifiable is the backing and the launch.
What AI Passport actually does
What AI Passport actually does is less about locking data away than about handing the user the controls. You bring your information together — connecting the apps and assistants you use every day, plus your own notes, links, files and thoughts — and then you see what your Passport knows about you. As you use apps and assistants, new information about you is created, and it goes straight into the Passport, where you can review it, edit it, and choose who gets to see it.
The permission model is the product. Give a supported tool — ChatGPT, Claude, Gemini, Perplexity, Spotify, Google Calendar, Gmail, Strava, Notion, Slack, and more than 800 connector sources through the developer platform — access to selected information for a period you choose, and revoke it whenever you want. Ask questions about your life and your data, and the Passport answers with references showing where each answer came from. “Think of your EGOIST AI Passport like camera permissions,” McGurk wrote on LinkedIn when the product launched. It is free forever to use.
Why it matters now
The bet Egoist is making is that personal context is about to become the most valuable — and most contested — layer of AI. Today, every assistant starts from zero and rebuilds its picture of you from whatever it can grab; tomorrow, the company argues, the picture should travel with you, under your control, from one assistant to the next. That is why the seed round's investor list reads like a statement of intent: Y Combinator and Goodwater for company-building, plus scientists from DeepMind and Meta who build the very systems the Passport is designed to sit between.
Whether a two-person company can set the terms for how AI handles personal data is an open question — and McGurk and Khachaturov are, by their own account, just getting started. What they have already done is turn a frustration shared on a Cambridge dance floor into a funded, launched product with a clear position: your context is yours, the tools ask permission, and you can take it back any time.
The playbook
- Start from a shared frustration, not a market map — McGurk and Khachaturov did not start with a market map. They started with a shared frustration — black-box assistants collecting personal data with no user control — and built the company around fixing it.
- Sell control, not just storage — AI Passport's pitch is a permission, not a vault: you decide what enters, what stays private, and what gets shared — “it’s not all or nothing, like it is today,” as McGurk put it at launch.
- Stay tiny, borrow credibility — Two people, one YC batch, a $3M seed from YC, Goodwater and scientists from DeepMind and Meta. A tiny team can carry a big idea when the backers are the right ones.
- Explain it with an analogy everyone already has — “Think of your EGOIST AI Passport like camera permissions,” McGurk wrote on LinkedIn at launch. A familiar analogy does more work than a paragraph of architecture.