I think a lot of it is marketing, to overemphasize the proposed capabilities.
Max Junestrand said that in Amsterdam on Tuesday. He was asked about the summer’s run of incidents in which AI agents hacked other companies. The warnings that followed deserve a pinch of salt, the Legora co-founder and chief executive told Politico, speaking at the HumanX conference.
Legora builds AI tools for lawyers. It is one of the largest AI companies in Europe. On the same day Junestrand spoke, Bloomberg reported that it is in talks to raise at least $300m at a pre-money valuation of about $8.5bn.
Europe should not try to build a frontier lab
Junestrand went further than the safety question. Europe should put no effort into building a frontier AI lab of its own, he said. He called the idea wishful thinking.
What it should do instead is build on top of the models it receives. Europe has an opportunity to lead the world in applications, he told Politico. The technology arrives as an input, and the value is in what gets built with it.
That is close to the line Ursula von der Leyen took in her State of the Union address. She backed an AI slowdown and argued Europe need not build frontier models to draw value from them.
Junestrand is unusually well placed to agree with her. Legora runs entirely on models it buys from American labs, and a European frontier lab would be a supplier rather than a competitor to it.
The position is also a commercial one. Every euro Europe does not spend building a lab is a euro available to the layer Legora occupies.
The part that costs him
He was less comfortable on what that dependence means in practice.
European customers have spent the past year waiting longer than American ones for the newest models from OpenAI and Anthropic. US firms and cybersecurity authorities have had first access. Politico asked whether the delay is a competitive disadvantage.
Of course it is, Junestrand said. Absolutely.
Politico describes the delays as repeated rather than occasional, and says the priority went to plugging cybersecurity gaps. It does not put a length on them, and neither does Junestrand. For a company selling to law firms that want the newest reasoning applied to a contract, any lag is a product gap rather than a procurement inconvenience.
He also said securing Legora against AI-driven attacks ranks above shipping new features, because the company holds legal data. Reaching the newest models quickly is part of how it does that. He described working with other companies to put them to use as they arrive.
So the same interview holds two positions. The capability warnings are overstated. Getting hold of those capabilities late is a real problem.
The timing, and the money
Bloomberg’s report landed a few hours after the interview. Rebecca Torrence wrote that Legora is in talks over a round of at least $300m, at roughly $8.5bn before the new money goes in.
Pre-money matters here. If the company raises $300m at that valuation, it closes above $8.8bn. Talks are continuing, terms could change, and Legora declined to comment.
That number sits against several others. Legora opened its Series D in March at $550m, valuing the company at $5.55bn. Six weeks later Nvidia and Atlassian joined a $50m extension, taking the round to $600m and the valuation to $5.6bn.
In August the company was reported to be chasing $10bn, barely a season after that. The figure Bloomberg describes now is well above where the last round closed and some way below what was reported in the summer.
Revenue has moved faster than any of it. The company reached $100m in revenue inside 18 months. Chief financial officer David Eckstein posted on Tuesday that it hit $200m in annual recurring revenue last week, and that the second $100m took under six months.
The numbers the company volunteered
He put several other numbers with it. Legora is on track to add more ARR this quarter than its entire revenue base at the end of 2025. Its pilot win rate is 75% so far this quarter. More than 40% of new business in the third quarter came from in-house legal teams rather than law firms.
The reach numbers are larger than the funding ones suggest. Eckstein said 130,000 lawyers use Legora each month, across 2,100 firms and legal teams in more than 80 countries. More than half of the AmLaw 50 are customers, and he named Salesforce, Palo Alto Networks, Baker McKenzie and White & Case among them.
He also said gross margin is positive and improved while revenue doubled. That is a pointed thing for a legal AI company to volunteer this month, and it is the company’s own figure rather than an audited one.
The software handles contract review, research and drafting, work that used to absorb junior lawyers’ hours. Legora describes itself as an agentic operating system for legal work, with its technical team in Sweden and its largest market now the United States. Iconiq, General Catalyst, Accel and Nvidia’s venture arm are among its backers. It started in 2023.
The company it is measured against
Harvey is the American legal AI company and Legora’s closest competitor. It announced a $550m round at a $15.6bn valuation earlier this month.
That gap is the frame investors use. A European company growing at Legora’s rate still carries roughly half its American rival’s valuation. The two sell to many of the same firms.
It is also the practical argument behind Junestrand’s position on frontier labs. Legora competes with Harvey on products, not on models, and both of them buy from the same suppliers.
What makes the remarks awkward
Junestrand made them on a day when 21 countries and the European Commission were doing the opposite.
The Dutch government published a joint statement on frontier AI on Tuesday. Ministers and leaders from 21 countries signed it, calling for mandatory testing before release. It floats a new international body to supervise the models, and its first signatory was Norway’s prime minister.
One of Europe’s largest AI companies spent the same day arguing that the alarm those governments are responding to is partly a marketing exercise.
Neither position is unreasonable on its own. They arrived within hours of each other, from the same continent, about the same models.
What the round will show is which reading the money agrees with. A company that has doubled revenue in under six months, selling software built on other people’s models, is asking investors to price the application layer at close to $9bn. That is a bet on exactly the division of labour Junestrand described in Amsterdam.