Health AI Weekly Roundup (9th June - 15th June )

DeepHealth (US): launched Reporting Pro, an AI radiology reporting tool that combines speech recognition, AI findings, measurements, draft impressions and structured reporting in one workflow. It is being deployed across RadNet, with external customers expected to go live next quarter. [Link]

Mosaic Clinical Technologies (US): launched Mosaic Reporting, an AI-native radiology reporting platform that builds reports in real time from natural speech. The tool sits inside MosaicOS and is already deployed across Radiology Partners-affiliated practices. [Link]

Subtle Medical (US): received FDA clearance for SubtleHD(CT), its first CT image-enhancement tool. The AI software reduces noise and improves low-contrast detectability across scanner generations. [Link]

DeepIntent (US): launched Helix AI, an agentic AI platform for healthcare marketers to query health and media data, build audiences and activate campaigns across DSP and social channels. [Link]

MHRA (UK): launched London Region I, a regulatory sandbox to test AI-enabled medical devices in live NHS settings under MHRA oversight. Up to 10 manufacturers will join the first phase, with expressions of interest opening in July 2026. [Link]

MHRA (UK): launched an AI sandbox to test tools that could predict medicines safety risks and side effects. Up to five AI approaches will be tested from summer 2026. [Link]

General-purpose LLMs vs clinical AI tools: a Nature Medicine study found GPT-5.2, Gemini 3.1 Pro and Claude Opus 4.6 outperformed OpenEvidence and UpToDate Expert AI across MedQA, HealthBench and 100 real clinician queries. The real-clinical-query benchmark used blinded reviews from 12 US clinicians. [Link]

Measuring Epistemic Resilience of LLMs Under Misleading Medical Context: an arXiv preprint introduced MedMisBench, a benchmark testing medical LLMs against misleading clinical context. Mean accuracy fell from 71.1% to 38.0%, with clinicians flagging serious potential harm in 38.2% of reviewed cases. [Link]

Rad AI & Yale New Haven Health System (US): expanded their partnership to deploy Rad AI’s generative AI radiology tools across Yale New Haven’s imaging network. The partners will also co-develop and clinically test new radiology technologies. [Link]

Abridge & NVIDIA (US): are developing a clinical-conversation AI model using NVIDIA’s Nemotron open-model family and Abridge’s de-identified clinical data. The model is expected later this year and will be used inside Abridge’s ambient documentation platform. [Link]

Kaopiz Holdings & QuantumTX (Vietnam/Singapore): signed an MOU to co-develop AI and IoT-powered preventive healthcare tools for ageing populations. The partnership will focus on predictive health insights, rehabilitation support and remote monitoring. [Link]

01Health (UK): raised $15 million to expand its specialist-care platform across the UK and US. The platform helps local clinics deliver services such as orthodontics and sleep medicine using AI patient engagement, workflows and specialist oversight. [Link]

Uncovr (France): raised $7 million to develop AI that turns surgical video into operative reports and coding suggestions. The seed round was led by Index Ventures. [Link]

OurMind (Netherlands): raised €2.1 million to expand its AI platform for reducing healthcare admin work. Its Notes tool converts consultation conversations into medical documentation and is used by 300+ general practices and 14 hospitals. [Link]

TurnUp (Belgium): raised €2 million to expand its AI platform for reducing patient no-shows in dental and medical practices. The platform predicts missed appointments and automates follow-up through its AI receptionist, Elissa. [Link]

Rylo (Israel): raised $85 million to expand its AI-powered communication tools for deaf and hard-of-hearing users. The round brings total funding to $101 million. [Link]