Botkin.AI: what the rise and shutdown of Russia's first registered medical AI teaches

• Росздравнадзор приостановил применение системы Botkin.AI — Коммерсантъ, 2023-11
• Росздравнадзор приостановил использование системы диагностики Botkin.AI — РБК, 2023-11-21
• Росздравнадзор приостановил применение Botkin.AI из-за отсутствия пострегистрационных отчётов (приказ №7880, приказ Минздрава №1113н) — Vademecum, 2023-11-21
• Разработчик Celsus приобрёл Botkin.AI — Vademecum, 2023-12-08
• Росздравнадзор снова разрешил использование Botkin.AI (корректирующие меры, экспертиза ВНИИИМТ) — Vademecum, 2024-05-17
• Цифровой врач: зачем продали Botkin.AI и о чём эта сделка говорит рынку ИИ в медицине — Forbes Russia, 2023-12
• Сколковский проект Botkin.AI получил первое в стране РУ на диагностическую платформу с ИИ (класс 2б, испытания в ЦКБ РАН) — Фонд «Сколково», 2020
• Botkin.AI интегрировался со столичным ЕРИС для поиска рака лёгкого на КТ-снимках (46 медучреждений, гранты Москвы 147,3 млн ₽) — Vademecum, 2020-08-19
• ИИ для медицины: Botkin.AI привлёк 100 млн рублей — Vademecum, 2019-05-29
• «Интеллоджик» привлекла 160 млн рублей (раунд B: Unicorn Capital Partners, «Ташир Медика»; 150 тыс. исследований в пилоте) — CNews, 2020-12-23
• Botkin.AI: прецедент или нет? (первая приостановка медизделия с ИИ, конец «песочницы») — HealthOps, 2023-11-21

Background

Botkin.AI was a medical-image analysis platform by Intellogic, founded in 2015 by Sergey Sorokin. A Skolkovo resident, the project became the flagship of Russian medical AI and, at the same time, its most instructive story: in eight years the company went from prototype to the country's first registered high-risk-class AI medical device — and then to the industry's first regulatory shutdown, worded as 'a threat of harm to citizens' lives and health'.

Money arrived in waves, and the investor roster shows how seriously the market believed in Botkin.AI. In 2017 — 11 million rubles from Primer Capital. In 2019 — 100 million rubles from Digital Evolution Ventures (a fund created with Rosatom's participation) and RBV Capital (linked to R-Pharm). In December 2020 — a 160 million ruble Series B: led by Unicorn Capital Partners, manager of the Industry Ministry's 'Pharmmed Innovations' fund, with Tashir Medika co-investing; earlier backers, including Primer Capital and Ekspocapital, joined as well. By the time of the sale, Tashir Medika held 17.6% and the founder retained 8.9% (Kommersant).

The regulatory trajectory looked exemplary. On November 6, 2020, Roszdravnadzor issued the platform a registration certificate as a class 2b (high-risk) medical device — at that point Russia's only such certificate for AI software of that class. The platform was registered for analyzing CT scans (malignant tumors, COVID-19), X-rays (tuberculosis, oncology, pneumonia), and mammography; clinical trials ran at the Central Clinical Hospital of the Russian Academy of Sciences. The company announced plans for international certification and expansion into Southeast Asia, Europe, and the U.S.

Technologically the platform relied, in the company's wording, on patented AI-based 'mathematical patient model' technologies and by late 2020 was integrated with 139 diagnostic devices (CNews). The industry backdrop was favorable too: from 2020 Moscow was rolling out its computer-vision experiment in radiology, and in February 2020 the city government approved 16 grants to medical software developers worth a combined 147.3 million rubles. Botkin.AI was among the main beneficiaries of that wave — and among its most visible faces: the project was backed by the Skolkovo Foundation, and its certificate was only the second in Roszdravnadzor's registry for AI software at all, and the first for medical image analysis (Vademecum). This case is not a success story: it examines why the market pioneer — with funding, registration, and government integration — was halted by the regulator and sold to a competitor.

Problem

The task Botkin.AI addressed was real and important. Lung cancer is often caught late, when treatment is far less effective, and AI analysis of chest CT promised earlier detection of nodules and fewer misses. The company publicly claimed roughly 95% accuracy on CT and 95–97% sensitivity for nodules of 6 mm and larger.

Deployment scaled quickly. In August 2020, Botkin.AI integrated with Moscow's ERIS service: CT studies from 46 of the city's medical facilities were automatically exported and analyzed by neural networks for signs of lung cancer. By the time of integration, 7,000 studies had been processed, with plans to reach 20,000+ per month. In the Moscow pilot from March 2 to December 7, 2020, the platform processed more than 150,000 studies and was integrated with 139 diagnostic devices. The system was also used outside the experiment — in the Otkrytaya Klinika chain, Moscow's telemedicine center, and Tula region projects.

But along with the class 2b certificate, the company took on an obligation that proved decisive. Under Health Ministry order No. 1113n (October 2020), manufacturers of high-risk medical devices (classes 2b and 3) must conduct post-registration monitoring of safety and clinical efficacy for three years and submit reports to Roszdravnadzor. In other words, the claimed 95–97% had to be demonstrated not once in trials, but systematically confirmed in real clinical operation — on a real patient flow, with real scanners and real physicians reading the outputs.

That is a fundamentally different mode of operation from showcasing a model on a labeled dataset: it requires processes for collecting feedback from clinics, staff to prepare the reporting, and a willingness to show the regulator inconvenient numbers. Botkin.AI's problem turned out to lie not in ambition or in the technology as such — but precisely in this verification loop, which never started working at a company with annual revenue of some fifteen million rubles.

Solution

The chronicle of the reversal — from industry showcase to a suspension order — can be reconstructed almost day by day from Kommersant, Vademecum, and RBC.

The rise: 2020 was the company's triumphant year. March–December — the Moscow pilot with 150,000 studies processed. August — the ERIS integration and status as one of the key services in Moscow's experiment: CT scans from 46 medical facilities flowed into the platform automatically. November — the country's first class 2b certificate for an AI platform; CEO Sergey Sorokin said the registration would 'open opportunities for using the platform in regional screening programs' (Skolkovo). December — the 160 million ruble Series B and a bid for international expansion: CE mark, Southeast Asia, Europe, the U.S. Sorokin framed the round's goal as a chance to 'solidify leadership on the Russian market and become a visible player on the international stage,' while Arkady Dvorkovich, speaking for Skolkovo, called the project a solution that 'saves lives' (CNews). Deployment also spread beyond the experiment: the Otkrytaya Klinika chain, Moscow's telemedicine center, and Tula region projects (Vademecum).

The problem builds: after registration, Roszdravnadzor — as order No. 1113n requires — repeatedly requested post-registration safety and clinical efficacy reports from Intellogic. Per Vademecum, 'despite repeated demands from Roszdravnadzor, the said reports from the Botkin.AI manufacturer never reached the service.' The regulator then ran its own monitoring — and found 'deviations in the characteristics (the AI software's outputs) relative to those declared at its state registration.'

The halt: on November 8, 2023, Roszdravnadzor issued order No. 7880 suspending the device's circulation 'due to a threat of harm to citizens' lives and health.' The agency's head, Alla Samoylova, publicly explained the decision at the Novamed forum: 'absence of clinical effect and repeatedly identified functional defects.' Moscow's health department confirmed Botkin.AI was no longer used in the city's clinics by then. Investor Unicorn Capital Partners insisted the system was undergoing 'technical reorganization' and that any threat of harm was 'completely excluded' — but a product officially labeled a threat to life cannot exist on the market.

The industry's reaction matched the moment: trade media dissected the case as a precedent — the first time Roszdravnadzor had ever suspended an AI medical device, which for the market meant the end of the 'sandbox' era AI services had enjoyed since the first registrations (HealthOps). Three theories circulated — from unfair competition to preemptive regulatory action — but the documented storyline came down to the most mundane one: the company ignored reporting requests, and the inspection found discrepancies with the registration dossier.

The endgame: in December 2023, 100% of Intellogic — with 15.5 million rubles in revenue and a 139 million ruble loss for 2022 — was bought by Medical Screening Systems LLC, owned by Igor and Vladimir Chernykh and developer of the competing Celsus AI platform (the change was entered into the state register on December 6). Artem Kapninsky, co-owner of the Celsus operator, took charge of the company. The buyer's own books put the deal in perspective: Medical Screening Systems' 2022 revenue was 1.6 million rubles.

An epilogue rarely remembered: in May 2024, Roszdravnadzor lifted the suspension. The new owner implemented corrective measures, which passed expert review by VNIIIMT — and the device formally returned to circulation. The technology proved fixable; by then, the business had already changed hands and ceased to exist as an independent company.

Result

The outcome cuts both ways. The product ceased to exist as an independent business: regulatory suspension in November 2023, sale to a competitor in December 2023, out of the public eye. Investors who had put in at least 271 million rubles across 2017–2020 exited against 15.5 million rubles of revenue and a 139 million ruble loss in the last full year. Yet AI radiology in Russia is alive and growing: Moscow's computer-vision experiment in radiology continues with dozens of AI services, and Intellogic's buyer keeps developing its own Celsus platform.

The framing matters. 'A threat of harm to life and health' is the legal basis of order No. 7880, not a proven fact of harm done: there is no public evidence of injured patients. Nor was the recall permanent — in May 2024, after corrective measures and VNIIIMT expert review, the suspension was lifted. And the cause of the halt was not 'the AI made mistakes' as such, but the failure to submit post-registration reports and the divergence of real-world characteristics from those declared at registration. It is also worth remembering that the parties' public statements contradicted each other — the regulator spoke of an 'absence of clinical effect' while the investor called the threat 'completely excluded' — and the market has no independent arbiter between those positions beyond the order itself and the subsequent VNIIIMT review.

In our view, the key lesson of this case is economic, not technological. Botkin.AI held a unique regulatory asset (the only class 2b certificate for an AI platform) and a government integration with the country's largest stream of studies — and still failed to build sustainable revenue: 15.5 million rubles a year did not cover even a fraction of the compliance costs a high-risk device class demands. Post-registration monitoring is a permanent expense line — clinicians, data, reporting; a company whose unit economics don't work cuts exactly that line, because the penalty is deferred. Here it arrived in its maximal form.

A second observation — also editorial: the real price of the failure was measured not in investors' money but in the evidence bar set for the whole industry. Since November 2023, every medical AI developer in Russia knows that a registration certificate is the start of regulatory work, not its finish line — and that the regulator can verify claimed metrics on its own. For the market's maturity, such a precedent may be worth more than another success story.

Lessons learned

1. A registration certificate is an obligation, not a trophy: Botkin.AI was halted not for a bad model but for missing post-registration reports and real-world performance diverging from declared figures.
2. For class 2b and 3 devices, Health Ministry order No. 1113n mandates three years of post-registration safety and efficacy monitoring — a permanent cost line that must be built into unit economics before obtaining the certificate.
3. Claimed metrics (95–97% sensitivity) must survive independent verification in operation — the regulator ran its own monitoring and found deviations; a marketing number turned into legal risk.
4. Medical AI economics are brutal: 15.5M RUB revenue against a 139M RUB loss in 2022 — even the leader with the first certificate, a government integration, and 271M+ RUB raised couldn't make unit economics work.
5. Regulatory death is not always final, but business moves faster than the regulatory cycle: the suspension was lifted within six months (after corrective measures and VNIIIMT review) — by which point the company had already been sold.
6. One player's failure doesn't bury the field: AI radiology in Russia kept growing after Botkin.AI's shutdown — but the evidence bar rose for everyone.
7. 'Move fast and break things' does not work in medicine: the cost of error is measured in the regulator's wording — 'a threat of harm to life and health'.

Frequently asked questions