AI already speeds up medical device development. The open question is how to use it inside a regulated environment without adding risk. That was the thread running through "AI and the Future of Medtech Innovation: From Idea to Impact", the ICI @ LSI panel at LSI Europe '26.
Rafael Beyar (Co-Founder and Managing Partner, ALIVE Israel HealthTech Fund) moderated. The panel brought together Daniel Adler (CEO and Co-Founder, BioT), Chen Levin (Co-Founder and CEO, Earendel Medical), Nitzan Hirsh (Co-Founder and CEO, LuSeed Vascular) and Yuval Shklarsh (Founder and CEO, Kudu Med).
Watch the full panel on the LSI site
The large majority of FDA-cleared AI-enabled devices are in radiology, with cardiology and neurology next. Radiology has DICOM and calibrated measurements. Surgery records very little usable data. Earendel Medical puts imaging sensors into single-use cannulas so surgery starts producing structured data.
At Kudu Med, a power-transfer design problem for an implant was turned into an AI-driven optimization problem with cost, manufacturability and durability constraints. Work estimated at about two years took about five weeks.
LuSeed Vascular's implant contains no AI, but the development process does. AI tools predict how each prototype behaves inside an aneurysm, next to bench and 3D-printed models.
BioT provides the cloud infrastructure for medical device companies. It secures data in transit and in the cloud, keeps patient information visible only to the relevant stakeholders, and routes data to clinicians, EHRs and third-party systems. About 40 customers use it to collect device-generated patient data and run algorithms and analytics on it, from hospital to home, without months of infrastructure development.
AI tools add new vulnerabilities and lower the skill needed to attack a system. AI assistants connected to company tools can take actions on their own. BioT keeps production systems separated and segmented from the systems that use AI.
The same question to an AI can return different answers. Validation requires consistent results, so guardrails on AI output are a must. Once security and consistency are solved, the software side of MedTech will move much faster.
Regulators, notified bodies and care providers set the requirements for moving device and patient data to the cloud. BioT holds ISO 27001, ISO 27799, SOC 2 Type II and HITRUST r2. SOC 2 Type II and HITRUST r2 require 18 to 24 months of documented operating history. A developer running many AI agents cannot review every line, and unreviewed code is hard to defend in an audit.
Earendel keeps its device disconnected until FDA clearance, to limit cloud security exposure at this stage.
Without control over the input data, a strong model can still mislead design, clinical and patient-selection decisions.
AI shortens development. It does not shorten the evidence trail that regulators and hospitals ask for. Teams that build on an audited platform layer can use AI in development and keep the infrastructure evidence in place. What stays with the device maker: the device, its clinical claims and its own submission.