Health AI Weekly Roundup (1st June - 08 June )

Kanza AI (US): launched its Clinical Reasoning System on Lightning AI’s NVIDIA infrastructure, with live use at Freyja Clinic and reported top ranking across 32 clinical reasoning dimensions in a 1,060-case evaluation. [Link]

NHS England & Microsoft (UK): announced a rollout of Microsoft 365 Copilot to 505,000 clinicians and support staff, after a 30,000-worker trial found average admin time savings of 43 minutes per staff member per day. [Link]

Genomics (UK): launched Mystra AI, an agentic human-genetics platform designed to support drug target discovery and validation using a data collection spanning more than 45,000 genome-wide association studies. [Link]

Clarius Mobile Health (Canada): launched its FDA-cleared Ejection Fraction AI for handheld ultrasound, automatically calculating and displaying cardiac ejection fraction in real time at the point of care. [Link]

Colorado HB26-1195 (US): signed into law to restrict AI use in psychotherapy, requiring licensed professional oversight and written consent when AI records or transcribes therapeutic sessions. [Link]

Colorado HB26-1139 (US): signed into law to require human review when AI is used in health insurance coverage decisions, including denials based on medical necessity. From January 2027, denials cannot be issued solely from AI output without review by a licensed clinician, physician, or other competent regulated professional. [Link]

European Commission CADA (EU): outlined a Cloud and AI Development Act proposal introducing an EU-wide sovereignty framework for cloud and AI, with public sector bodies using risk assessments to choose assurance levels for cloud providers. [Link]

A large-scale vision foundation model for musculoskeletal radiographs: researchers introduced SKELEX, a foundation model trained on 1.2 million MSK X-rays and evaluated across 12 diagnostic tasks, showing broad performance and anomaly localization without task-specific training. [Link]

Gender-Dependent Diagnostic Substitution in LLM Medical Triage: Same Symptoms, Unequal Urgency: an arXiv study found that three LLMs gave young women lower ER referral rates than age-matched men with identical neurological symptoms, linked to models anchoring more often on idiopathic intracranial hypertension in young women. [Link]

Mayo Clinic & Microsoft (US): partnered to develop a healthcare-focused frontier AI model using Mayo Clinic’s de-identified clinical data and Microsoft’s AI, cloud, and engineering capabilities. [Link]

Sanofi (France) & Owkin (US/France): expanded their partnership through a multi-year collaboration to build AI agents for drug research and development, deployed through Owkin’s K Pro platform. [Link]

SOPHiA GENETICS (Switzerland) & MSK (US): signed an MOU to explore a joint precision medicine hub combining MSK’s oncology data and expertise with SOPHiA GENETICS’ AI analytics platform. [Link]

Alnylam & Inceptive (US): formed an AI collaboration worth up to $2 billion to accelerate RNAi therapeutic discovery, combining Alnylam’s siRNA data with Inceptive’s biological foundation models. [Link]

Collate (US): raised $95 million in funding led by Redpoint Ventures to scale its AI tools for automating regulatory, clinical trial, and product development documentation in life sciences. [Link]

Stepful (US): raised $55 million in Series C funding led by Oak HC/FT to expand its AI-assisted healthcare workforce training platform for hospitals. [Link]

Adaptive Innovations (US): announced $60 million across Series A and Seed funding to expand its AI-native home health provider model, combining in-home clinicians with AI operations for intake, scheduling, charting, billing, and compliance. [Link]

Flok Health (UK): raised $12.5 million in Series A funding led by AlbionVC to scale its AI-led physiotherapy clinic, currently available to 2.4 million NHS patients across 11 areas. [Link]

Elation Health (US): acquired Aster, an AI-native women’s health EHR startup, to expand its agentic AI capabilities for primary care. [Link]

Elsevier (Netherlands): acquired Wellsheet to bring EHR patient data into ClinicalKey AI, delivering patient-specific clinical evidence directly inside existing workflows. [Link]