DeepL just landed one of the more interesting B2B deals in the AI translation space. The Cologne-based company has struck a partnership with Harvey, the legal AI startup valued at $11bn and reportedly in talks to raise at a $15.5bn valuation, to handle document translation across Harvey’s platform. DeepL will cover more than a third of Harvey’s total document translation volume. That’s a meaningful share, and a clear signal about where DeepL thinks its future is.
The deal makes strategic sense on both sides. Harvey serves over 2,000 lawyers across more than 2,400 organizations, and legal work is inherently cross-border. Contracts, filings, briefs, evidence packages, client communications — all of it needs to move between languages quickly and accurately, often at high volume. Generic translation tools struggle with that. Legal terminology is precise, context-dependent, and unforgiving. A mistranslation in a contract clause is not a minor UX problem.
For DeepL, this is part of a deliberate pivot toward regulated industries. The company has been explicitly targeting legal, financial services, pharmaceuticals, and life sciences, arguing that its accuracy and security standards are built for environments where errors carry real consequences. Its existing legal clients include law firm Taylor Wessing. Other clients across industries include SoftBank and Mazda. So the Harvey deal adds weight to a thesis DeepL has been building for a while.
Still, context matters here. Earlier this year, DeepL cut 250 employees, which raised questions about the company’s trajectory. Doubling down on enterprise deals in high-value verticals looks like the response to that. Rather than competing on breadth against tools like Google Translate or DeepL’s more consumer-facing rivals, the company is narrowing its focus to sectors where specialization commands a premium.
Harvey itself is not without competition. Swedish startup Legora is one direct rival in the legal AI space, and the broader legal tech market has attracted significant investment from multiple directions. But Harvey’s scale, at over 2,400 organizations, gives DeepL a substantial distribution channel into exactly the buyer profile it wants to reach.
The partnership also fits a broader pattern in AI right now: vertical-specific stacks built from best-in-class components rather than single-vendor solutions. Harvey handles the legal reasoning and workflow layer. DeepL handles translation. That kind of composable approach is increasingly how serious enterprise AI gets deployed, and it’s a smarter architecture than trying to build everything in-house.
For developers and founders watching the AI infrastructure space, the more interesting question is whether DeepL can hold its position as the default translation layer in legal and regulated industries, or whether larger players move into that gap. For now, the Harvey deal gives it a strong anchor.
