Biolevate announces a €30 million Series A round to deploy its artificial intelligence in regulated operations in healthcare industries. The Parisian company no longer presents itself only as a medical writing tool, it wants to organize the way in which laboratories research, document, verify and validate their scientific knowledge. An ambition which places it in the face of established regulatory platforms, but also at the heart of the question of the sovereignty of the uses of AI in health.
Artificial intelligence promises to shorten the time between a scientific discovery and the arrival of a treatment on the market. In laboratories, it already makes it possible to browse publications, reconcile data sets, prepare first drafts of documents or identify signals in regulatory monitoring. But, in this industry, producing a response is only the beginning of the job. It is still necessary to be able to establish the source, understand the method used, verify the data used and, above all, attribute the decision to an identifiable expert.
It is on this last part, less visible than the discovery of molecules but decisive for the clinical development and authorization of drugs, that Biolevate wants to position itself. The startup announces a Series A of 30 million euros, co-led by RAISE France and Orange Ventures, with the participation of EQT Ventures, the MSD Global Health Innovation Fund and Station F. It had raised 6 million euros in seed in November 2024.
The funding is expected to accelerate product development and international expansion, with the opening of an office in Boston. Aymar Hénin, entrepreneur and investor via RAISE France, takes over as chairman of the board of directors; Jérôme Berger, director of strategy and venture capital at Orange, is also included. The composition of the round gives an indication of the stage which is opening; it is no longer just a question of financing a technology, but of installing it among large health groups.
The bottleneck is after discovery
AI applied to health is often told from upstream: better identification of therapeutic targets, molecules designed more quickly, better selected clinical trials. This part of the chain concentrates the announcements and a large part of the capital, but the frictions continue well after the research laboratory.
A therapy must be documented in thousands of pages of publications, protocols, trial reports, responses to authorities and submission files. Regulatory teams must follow different rules depending on the country, ensure the consistency of documents produced over several years and demonstrate that each conclusion is supported. The potential time savings are considerable, and so are the risks of poorly controlled automation.
Biolevate first approached this market through medical writing, that is to say the preparation of scientific and regulatory documents. The startup is now expanding its focus to research, clinical trials, health technology assessments, regulatory files and drug development. Biolevate no longer seeks only to help an editor go faster, but to circulate controlled knowledge between several functions in the laboratory.
Agents, but above all a chain of proof
The platform boasts a knowledge-based approach by aggregating scientific, regulatory and company-specific sources, then using them to power workflows in which results remain linked to the documents that support them. Agents can search, extract, compare and synthesize, while experts maintain control over review and validation.
In a regulated environment, the issue is not just about reducing erroneous responses from a model, but it is about reconstructing the complete path from a source to an assertion, from an assertion to a document, and then from a document to a decision. An unsourced answer can be embarrassing in a search engine; it becomes an operational risk when it enters a file intended for a health authority.
In terms of performance, Biolevate advances literature reviews carried out up to sixteen times faster and 80% efficiency gains in regulatory monitoring and compliance, and communicates on the capacity to operate 100,000 agents simultaneously.
Customers in production, the real signal of lifting
Biolevate also indicates that it has more than ten major account customers in production, after having established partnerships with Sanofi, NVIDIA and Microsoft. It is these deployments which give the operation its scope, because if in health, a pilot is easy to launch, often financed by innovation or digital management, a use put into production supposes a completely different discussion: data security, qualification of suppliers, integration with existing tools, quality control and responsibility of business teams.
The startup now has 50 employees, and plans to double its workforce again.
Finally, Biolevate has also initiated seven patent applications, two of which have already been filed in Europe.
The competition already has either the files or the teams
Biolevate is entering territory that is far from virgin, in medical and regulatory writing, Yseop, Certara and Indegene have already built offers for automation, document generation and support for writing teams. Yseop, also French, is the most direct comparable on the promise of traceable AI capable of producing clinical and regulatory documents.
Further up the chain, Veeva and ArisGlobal have a structural advantage: their platforms already manage, for many laboratories, product information, document versions, validations and exchanges with authorities. Both are now adding GenAI and agent capabilities to their own environments. Clarivate, for its part, holds regulatory and scientific intelligence bases that laboratories use to make their decisions.
Finally, large service providers, from Indegene to IQVIA, own the teams that still perform a considerable portion of the editorial, vigilance, and regulatory affairs work. Their challenge is simple: transform expertise sold in man-days into software workflows that they control themselves.
Biolevate’s place will therefore not be to replace all these players alone. It must demonstrate that it can become the layer that connects knowledge to operations, on top of existing reference systems and without creating an additional tool that teams must administer. This is a harder value proposition to sell than an editorial assistant, it is also, if it works, a much more defensible position.
A strategic startup for France, under certain conditions
The strategic nature of Biolevate does not come from the fact that it uses AI in health. France already has startups in drug discovery, medical imaging and hospital data. It lies in the location of the value chain it targets: the organization and validation of sensitive knowledge within industries that are largely dependent on foreign software, data and models.
Such a platform can give European laboratories the possibility of deploying agents on their own corpora, in controlled environments, with an explicit policy on sources, data and people who validate. The issue of sovereignty is therefore not limited to the choice of a language model, it concerns the application layer where business rules, access rights, control procedures and decision memory are included.
This qualification as a strategic startup, however, remains conditional. It assumes that Biolevate maintains and deepens its mastery of workflows, evaluation and integration, rather than becoming an interchangeable interface on top of models designed elsewhere. It also assumes that the decision center, the intellectual property and the capacity to serve European customers remain anchored in France and Europe, while allowing the company to win the American market.
Boston is, from this point of view, an almost obligatory stopover. The city concentrates the laboratories, biotechs and regulatory skills that can make Biolevate a global supplier. The difficulty lies in developing there without the company’s center of gravity leaving Paris, like so many young European startups when large-scale marketing begins.
An open trajectory between independence and consolidation
This Series A does not open an imminent exit scenario, and gives Biolevate the means to try to establish itself as an independent platform. To achieve this, the company will have to convert its first deployments into lasting contracts, strengthen its integrations and show that its technology applies to several therapeutic areas and regulations without multiplying tailor-made projects.
If this trajectory is successful, several families of buyers could come forward. Regulatory software providers would benefit from integrating a more advanced layer of reasoning and proof. Service and CRO groups could seek to industrialize part of their operations. A supplier of models or infrastructure could also see Biolevate as a gateway to a regulated vertical, provided that the technological neutrality required by laboratories is preserved.
The latter assumption will depend not only on the performance of the models, but on Biolevate’s ability to become a difficult-to-rebuild asset: a team that understands regulatory affairs, customers who actually use the platform, and processes whose results are accepted by quality functions.
