DeepL Powers a Third of Harvey Legal Translations After Pivot

DeepL will handle more than a third of all document translations inside Harvey’s legal AI platform after the Cologne company embedded its models directly into the US startup’s workflow. The move lands three months after DeepL cut 250 roles and locked its focus on regulated sectors.

Harvey, already used by more than 200,000 lawyers across 2,400-plus organisations in 70 countries, treats translation of contracts, filings, briefs and evidence as one of its heaviest daily jobs. DeepL now sits inside that loop via its API.

The arrangement is not a side plug-in. It places a specialised language layer under one of the busiest workflows in enterprise legal software, at a moment when both companies are reshaping how they grow.

Harvey Puts DeepL Inside the Platform

Customers stay inside Harvey. They upload any document, pick a target language from more than 100 options, and receive a translation that keeps original structure, formatting and approved terminology. No export, no reformatting.

Lauren Oh, product manager at Harvey, said document translation ranks among the platform’s most-used workflows. “By integrating DeepL’s AI document translation capabilities directly into the Harvey platform, we’re giving legal teams an even faster, more seamless and reliable way to work across languages, while also making sure content stays highly precise and accurate.”

Working with Harvey brings our specialized Language AI capabilities to even more legal teams around the world, who work across borders every day. Legal work usually means dealing with a lot of documentation, whether that’s contracts, filings or case documents that span hundreds of pages, in dozens of different formats and unique contexts.

Jarek Kutylowski, Founder and CEO of DeepL

DeepL is one of several translation vendors Harvey uses. The German firm was chosen for context-aware models, custom glossaries that lock firm-approved terms, and broad file-format support that keeps long, complex packs intact.

That combination matters because legal packs rarely arrive as clean plain text. Clause numbering, defined-term tables, exhibits and bilingual schedules all have to survive the round trip. Keeping lawyers inside a single interface removes the hand-offs that usually introduce version drift.

Over a Third of the Translation Load

The volume share is the concrete stake. DeepL will process over a third of Harvey’s total volume of document translation. That figure sits on top of Harvey’s existing customer base and DeepL’s own enterprise footprint of more than 200,000 business teams.

Harvey closed a $200 million raise at an $11 billion valuation in March 2026, co-led by GIC and Sequoia. Reports in early August put it in talks for at least $500 million at $15.5 billion. DeepL reached a $2 billion valuation in 2024 after a $300 million round led by Index Ventures.

The valuation gap underlines different stages. Harvey is still scaling agents and seats at a rapid clip. DeepL is converting an already large enterprise base into deeper, higher-trust usage inside vertical platforms. A third of Harvey’s translation traffic is recurring volume tied to matters that already clear legal procurement, not one-off consumer queries.

The 250-Job Pivot That Cleared the Path

In May 2026 DeepL reduced headcount by about 250 roles, roughly a fifth to a quarter of staff at the time. CEO Kutylowski framed the cuts as a structural choice for an AI-native company: smaller teams, fewer layers, heavier use of its own models.

That same period saw DeepL double down on heavily regulated verticals: legal, financial services, pharmaceuticals and life sciences. The company already counted Taylor Wessing among legal clients and SoftBank and Mazda among broader enterprise names. The Harvey integration is the clearest proof point yet that the narrower focus is producing volume.

Readers following the earlier cuts can see the through-line in our coverage of DeepL’s 250-job AI-native restructuring. The partnership announcement three months later shows the strategy landing inside a high-growth platform rather than competing as a pure end-user tool.

The sequence is tight enough to read as a single arc rather than separate news cycles:

1. May 2024 – DeepL raises $300 million at a $2 billion valuation.
2. March 2026 – Harvey closes $200 million at an $11 billion valuation.
3. May 2026 – DeepL cuts about 250 roles and narrows on regulated sectors.
4. Early August 2026 – Reports put Harvey in talks for at least $500 million at $15.5 billion.
5. 19 August 2026 – Harvey announces the DeepL integration and the one-third volume share.

Three months from restructuring announcement to a flagship legal distribution deal is short. It suggests the vertical push was already under way when the headcount change landed, not a late scramble for relevance.

Why Cross-Border Legal Work Needs This Stack

Cross-border matters generate contracts, evidence packs, regulatory filings and client reports that run hundreds of pages and switch languages mid-matter. Accuracy on defined terms, preservation of clause numbering and formatting, and audit-ready security are non-negotiable.

DeepL’s enterprise stack meets those bars with:

• No permanent data retention for enterprise and paid users
• GDPR compliance plus SOC 2 Type II and ISO 27001
• Custom glossaries that enforce firm or client terminology
• Native support for complex file formats without layout loss

Harvey’s product team can therefore keep lawyers inside one interface while the specialised translation layer runs underneath. The same security posture that DeepL markets for Language AI tools built for legal work also covers finance and life-sciences documents.

In practice the stack splits labour. Harvey owns matter context, user permissions and agent workflows. DeepL owns language fidelity, glossary control and format survival. Neither side has to rebuild the other’s core strength to ship a usable multilingual path.

Legal AI’s Two Heavyweights and the Swedish Rival

Harvey sits at the top of the enterprise legal-AI market alongside Swedish rival Legora (formerly Leya). Both target Am Law 100 firms and large in-house teams with agents that handle diligence, contract review, research and multi-step workflows. Pricing for Harvey starts in the enterprise range with multi-seat minimums.

Translation is not the core product for either platform. It is infrastructure. By routing a third of that load to DeepL, Harvey avoids building and maintaining specialised language models while still offering 100-plus languages with legal-grade fidelity. Legora and other competitors will face the same pressure to offer seamless multilingual support as cross-border matters grow.

DeepL, for its part, gains distribution that standalone marketing cannot match: every new Harvey seat that touches a foreign-language document becomes a DeepL transaction.

That distribution logic cuts both ways. Harvey can market multilingual coverage without carrying the full cost of language R&D. DeepL reaches lawyers who already cleared security review for a major legal AI platform, shortening the usual enterprise sales cycle for language tools.

Regulated Verticals Beyond Contracts

The deal is one data point in a wider shift. DeepL now pitches the same accuracy-and-security combination to financial services and pharma teams that handle clinical reports, regulatory submissions and multi-jurisdictional filings. Nearly half of the Fortune 500 already use DeepL in some form; the legal win supplies a reference case for other compliance-heavy buyers.

Voice-to-voice translation and real-time meeting features extend the same models into spoken work, but document volume remains the immediate revenue driver inside Harvey. Kutylowski has repeatedly described Language AI as core infrastructure for global business. Embedding inside vertical platforms is how that claim turns into recurring, high-trust usage.

For now the numbers are clear: one integration, more than a third of Harvey’s translation traffic, and a post-restructuring DeepL that is smaller, more focused, and sitting inside the dominant legal AI system.

Legal is the loudest proof point today because contract and filing volume is so visible. The same retention, certification and glossary controls travel cleanly into adjacent regulated work once a buyer sees them operating at scale inside a peer platform.

How Lawyers Move Through the New Path

The daily path is deliberately short. A lawyer opens a matter in Harvey, uploads a contract pack or evidence set, chooses a target language from the hundred-plus list, and receives a formatted return without leaving the platform.

Behind that click, DeepL’s models apply firm or client glossaries, hold structure across long files, and operate under the no-permanent-retention rule that enterprise accounts require. Harvey continues to manage access control, matter context and any downstream agent steps that use the translated text.

Because DeepL is one of several translation vendors inside Harvey, the platform can still route other work elsewhere. The one-third share simply shows where volume is concentrating after the direct embed. Specialised packs that need locked terminology and intact layouts are the natural fit for the DeepL path.

The design also limits training burden. Teams already cleared on Harvey do not learn a second interface or a separate export routine. That lowers friction on matters that switch language mid-stream, which is common once filings, local counsel drafts and client reports enter the same file set.

What the Traffic Split Means Going Forward

A fixed share of a growing platform’s translation load is more durable than a standalone seat count. Harvey’s lawyer base already more than doubled earlier in 2026, from roughly 100,000 lawyers across 1,300 organisations to more than 200,000 lawyers across 2,400-plus organisations in 70 countries. Each new seat that touches foreign-language material can feed the same API path.

For DeepL, the deal converts the post-cut focus on regulated work into measurable throughput inside a category leader. Taylor Wessing already sat on the legal client list; Harvey multiplies reach across Am Law 100 firms and large in-house teams without DeepL having to win each firm as a separate end-user sale.

For Harvey, outsourcing a third of translation volume preserves engineering attention for agents, diligence flows and multi-step review. Multilingual coverage stays a platform feature rather than a parallel model-training programme.

Competitors watching Legora and the wider legal AI field now see a clear benchmark: seamless language support is table stakes for cross-border matters, and specialised partners are a viable way to deliver it. The next pressure point is whether other platforms match the same depth of glossary control, format survival and security certifications that this stack already advertises.

Frequently Asked Questions

How much of Harvey’s translation volume will DeepL handle?

DeepL will process over a third of Harvey’s total document translation volume through the direct API integration announced on 19 August 2026. The share covers contracts, filings, briefs, evidence and related materials uploaded inside the Harvey platform.

What security certifications does DeepL bring to legal work?

Enterprise and paid DeepL accounts use no permanent data retention, plus GDPR compliance, SOC 2 Type II and ISO 27001. Those controls, along with audit logs and SSO options, match the confidentiality requirements Harvey sets for high-stakes legal materials.

When did DeepL last raise money and at what valuation?

DeepL raised $300 million in May 2024 in a round led by Index Ventures that valued the company at $2 billion. That figure remains the most recent public valuation cited in coverage of the Harvey deal.

Who are Harvey’s main competitors in legal AI?

Swedish firm Legora is widely viewed as Harvey’s closest rival for large law firms and corporate legal teams. Both platforms emphasise agents for diligence, contracts and multi-step workflows; translation is typically supplied by specialised partners rather than built in-house.

How many lawyers currently use Harvey?

Harvey reports more than 200,000 lawyers across over 2,400 organisations in 70 countries as of the August 2026 announcement. Earlier 2026 figures had cited roughly 100,000 lawyers across 1,300 organisations, reflecting rapid growth through the year.