Scarlet, a London company that checks whether medical software and artificial intelligence (AI) tools are safe to put in front of patients, has raised $50M in a Series B round led by the venture capital firm Headline. The startup is one of the bodies European and British regulators have authorised to approve medical devices, and it focuses only on software and AI.
Notion Capital joined as a new investor, alongside existing backers Atomico, Creandum, Kindred Capital and Entrepreneurs First. Nancy Xiao McLaughlin, a partner at Headline, will join Scarlet’s board.
“Scarlet helps regulation work with and for, not against, innovation,” she said. The company will use the money to expand in the United States (US) and Asia-Pacific, add new types of certification and grow its team of AI reviewers and engineers.
The Gatekeeper For Health Software
In Europe, most medical devices cannot go on sale until an independent organisation called a notified body has checked them against the European Union’s (EU) Medical Device Regulation (MDR). The rules cover everything from pacemakers to apps that read X-rays or spot skin cancer.
A notified body reviews a device’s technical file, including its clinical evidence and the maker’s quality processes, before issuing the certificate that lets the device carry the CE (Conformité Européenne) mark used across the EU. There are about 52 such bodies under the MDR, and Scarlet says it is the only one that specialises in software and AI.
Britain runs its own version of the system. In January 2024, the Medicines and Healthcare products Regulatory Agency (MHRA), the United Kingdom’s (UK) regulator, designated Scarlet as an Approved Body to certify software and AI as a medical device. That took the number of active UK Approved Bodies to nine, and the MHRA said the new designations were meant to increase the capacity to certify the safety and performance of medical devices.
Scarlet, founded in 2021, spent its first three years working with regulators to obtain its designation before it issued a single certificate, because a certification body must itself be approved before it can approve anyone else’s product. Scarlet is also recognised by the US Food and Drug Administration (FDA) as a third-party review organisation, one of the outside groups the agency allows to review certain applications on its behalf.
Why Approvals Take So Long
Scarlet’s pitch rests on a mismatch between how fast software is built and how slowly it is checked. A company can train an AI diagnostic model in a few weeks, then wait 12 to 18 months for an expert assessment before it can sell it. Approval for even a minor update can add months of delay.
Scarlet’s founders, James Dewar and Jamie Cox, both built AI medical devices at Babylon Health. Babylon, a British digital health company, offered AI symptom checking and online doctor appointments, raised $550M in a single round in 2019 and collapsed in 2023.
“For every day that safe and effective medical technology is not available, someone’s life takes a different course,” said Dewar, Scarlet’s co-founder and chief executive.
How Scarlet Checks A Product
Customers upload their technical documentation, in any format, to Scarlet’s AI-enabled online platform. Qualified human reviewers then assess the submission against the relevant regulations and issue a certification decision. Customers pay a subscription that covers annual audits, reviews of later product changes and ongoing surveillance once a product is on sale.
In 2024, Dewar told the business magazine Fortune that approvals for updates to previously approved software, which once took six to eight months, could now take days. “We’re building AI-native certification to help our customers bring cutting-edge innovation to patients around the world,” he said of the new round.
By 2024, the FDA had approved close to 1,000 AI-enabled medical devices, and about 97% of them went through the 510(k) route, which lets a product reach the market by showing it is similar to one already approved, Fortune reported. A study in the journal Nature Medicine found that 43% of FDA-approved AI devices lacked clinical validation data, and that of 521 AI devices examined, 22 had results from randomised controlled trials.
“There’s an element of buyer beware,” Dewar told Fortune, adding: “At the moment, the regulation doesn’t ask us to do anything about bias.”
Who Uses It
Scarlet’s customers include Flo Health, the period- and fertility-tracking app with more than 500M downloads, and DeepHealth, a US radiology AI company. Others are Skin Analytics, whose software helps detect skin cancer, and NHS England, which runs the National Health Service (NHS) in England. The AI scribe makers Heidi Health and Tandem, whose tools write up doctors’ notes, are also customers.
Finn Stevenson, co-founder and chief executive of Flok Health, which runs an AI physiotherapy service used by the NHS, said: “Scarlet has built a Notified Body for the AI world.” Scarlet has not said how many certificates it has issued so far. It raised a $17.5M Series A led by Atomico in 2024, and the new round is nearly three times that size. Investors are also backing other healthcare AI companies, and Morning Tick has reported on Vocca, which raised $20M for AI that answers patient calls.
Faster Checks, Same Rules
Scarlet does not change the rules medical AI must meet; it changes how quickly companies can show they meet them. With designations in the EU and the UK and recognition from the FDA, it can offer certification and review services across all three markets.
The new money will test whether a certification body built around software can keep pace as more AI tools head for clinics and patients’ phones. For developers who have waited a year or more for approval, how quickly Scarlet clears each new version is likely to be the clearest measure of whether its model scales.
