Clinical & Diagnostics
6 Health Systems Enhancing Care Delivery with Ambient AI Scribes
The multicenter JAMA study spanned five academic medical centers tracking pre/post ambient scribe deployment metrics. Systems reviewed are using Nuance DAX, Abridge, and Suki as their core ambient NLP platforms. The JAMA Network Open companion study measured burnout via validated Maslach Inventory scoring and found a 74% reduction in burnout odds at the 30-day mark — the most rigorous clinician wellness data published to date on this category.
Amazon Connect Health Brings Agentic AI to the Point of Care
Amazon Connect Health integrates agentic AI agents with EHR data streams via FHIR APIs for real-time context retrieval during patient encounters. The architecture uses AWS Bedrock foundation models combined with Connect's contact center infrastructure to support intelligent call routing, automated prior authorization lookups, and proactive care gap alerts. HIPAA-compliant data handling is enforced at the infrastructure layer, with role-based access and full audit trails.
Americans May Be Losing Trust in AI in Health Care
The survey, fielded in March 2026 with a nationally representative sample, found the steepest trust decline among patients 55 and older, where acceptance fell from 45% to 31%. Concerns center on AI making diagnostic errors without physician oversight (67%), data privacy (61%), and loss of human connection (58%). Only 27% of respondents said they had received any explanation from their provider about how AI tools were being used in their care.
WHO Europe Publishes First Snapshot of AI in Health Care Across EU Member States
The report surveyed all 27 EU member states plus Norway, Iceland, and Liechtenstein. AI-assisted diagnostic tools are most prevalent in radiology (deployed in 21 of 27 states) and pathology (14 of 27). Patient chatbot deployments are primarily triage and appointment scheduling. Only 9 states have published formal AI governance policies for clinical settings, and the report flags wide disparities in AI infrastructure between high- and low-income EU member states as a major equity risk.
Research & Science
Clinical AI Has Boomed — Stanford-Harvard Report Shows What Actually Holds Up in Practice
The ARISE report analyzed 200+ clinical AI studies and found multi-agent frameworks delivered diagnostic accuracy gains of 7% to over 60% versus single-agent baselines depending on domain. Top failure modes in real-world settings include distribution shift (academic-trained models underperforming at community hospitals), annotation noise in training labels, and insufficient interpretability for high-stakes decisions. The report calls for mandatory prospective validation before clinical deployment and proposes a new evidence tiering framework for clinical AI.
How to Meaningfully Evaluate AI in Clinical Medicine
The proposed framework distinguishes between technical performance (accuracy on holdout sets), clinical utility (measurable improvement in patient outcomes), deployment robustness (performance stability across sites and populations), safety (adverse event tracking), and equity (performance parity across demographic subgroups). The authors reviewed 312 clinical AI validation studies published in 2023–2025 and found only 14 met all five criteria. Prospective, site-diverse, outcome-linked trials are presented as the minimum standard for high-risk clinical AI.
Precision Oncology in the Age of AI: Lessons from AI-Driven Drug Discovery and Clinical Translation
The review covers AI-guided target identification, generative molecular design using transformer and diffusion architectures, and multi-omics integration combining genomic, proteomic, and transcriptomic data streams. Key progress includes identifying actionable alterations in 40% of tumors previously classified as having no targetable mutation, and generative models producing novel small-molecule candidates for KRAS G12D — a mutation implicated in ~25% of all human cancers. Phase I trials for three AI-designed oncology compounds are underway as of Q1 2026.
Policy & Regulation
Aidoc Wins FDA Clearance for Comprehensive Foundation Model AI Covering 14 Indications
The CARE model achieved FDA clearance under a novel "platform authorization" pathway that evaluates the underlying foundation model rather than individual indication-specific models. Clinical validation data covered 11 newly cleared indications added to 3 previously cleared ones, with pooled sensitivity and specificity data across 150,000+ cases. The clearance sets a precedent for foundation model evaluation that both FDA and industry are expected to build on throughout 2026 as multi-indication AI systems proliferate.
FDA Cuts Red Tape on Clinical Decision Support Software and Wearables
The guidance clarifies that software which presents information to clinicians who independently review and act on it — rather than taking autonomous action — does not meet the definition of a medical device under the FDCA. This excludes large classes of AI-powered clinical summary, documentation assistance, and risk-scoring tools from the 510(k) clearance pathway. Most high-risk AI obligations under the revised framework take effect in August 2026, with full compliance for medical device AI required by August 2027.
TEFCA Reaches 500 Million Health Records Exchanged as HHS Leverages AI for Interoperability
TEFCA operates via Qualified Health Information Networks (QHINs) and uses FHIR R4 as the standard exchange format. The 500M record milestone includes clinical summaries, medication histories, lab results, and imaging metadata. HHS has announced draft USCDI v7 adding 29 new data elements to strengthen AI-readiness of exchanged data, including adverse event reporting fields, nutrition data, and expanded social determinants of health structured data elements.
Industry & Business
Digital Health Funding Hits $7.4B in Q1 2026 Driven by AI Drug Discovery and M&A
Q1 2026's 19 mega-rounds (deals of $100M+) accounted for 60% of all capital raised in the quarter. Top deals include Abridge's $300M Series E at a $5B valuation, Ambiance's $243M Series C at $1.04B, and Function Health's $300M Series C at $2.2B. On the M&A side, DeepHealth's $269M acquisition of Gleamer was driven by a combined footprint of 700+ hospital contracts. Non-clinical workflow automation, clinical workflow tools, and data infrastructure were the three highest-funded sub-categories.
Merck and Mayo Clinic Announce AI-Enabled Drug Discovery and Precision Medicine Collaboration
The collaboration will apply AI-guided target identification using Mayo's multimodal data — integrating genomic sequencing, proteomics, clinical histories, and imaging — to surface drug-disease associations not visible in isolated data types. Merck brings its AI-powered molecular design capabilities, including generative models for small-molecule synthesis optimization. The deal structure includes joint IP ownership on discoveries emerging from the collaboration, with clinical trial design for validated candidates a stated deliverable.
Jimini Health Raises $17M Seed to Launch AI Mental Health Platform Sage
Sage is designed for patients with complex mental health needs — bipolar disorder, PTSD, schizophrenia — a population that consumer AI chatbots have largely avoided due to clinical risk. The platform combines large language model-driven conversational support with clinician supervision workflows that flag high-risk conversations for human review. The $17M seed is led by General Catalyst, with participation from ARCH Venture Partners. Enterprise pilots with three large behavioral health organizations are planned for H2 2026.
UnitedHealth Group Is Making a $3 Billion Bet on AI — What Does It Mean for Patients?
The $3B commitment includes $1.2B in new AI infrastructure, $900M in model development and licensing, and $900M in workforce retraining and change management. UnitedHealth's prior authorization AI — already processing 80% of routine authorization requests without human review — is a focal point for regulators after a 2024 Senate investigation found denial rates for AI-processed claims running 22% higher than human-reviewed claims. The company disputes that figure and says its AI reduces incorrect denials.
Social Buzz
AI in the Mental Health Care Workforce Is Met With Fear, Pushback — and Enthusiasm
The NPR piece surfaces a fault line that LinkedIn and mental health forums have been debating all spring: AI administrative tools (scheduling, billing, session summaries) have near-universal enthusiasm from small practices drowning in paperwork, while AI clinical tools (risk scoring, diagnosis support, chatbots for between-session support) face deep skepticism. The most-shared critique: mental health documentation AI is trained primarily on structured clinical notes, not the nuanced, relational language that defines therapeutic encounters — leading to summaries that are technically accurate but clinically hollow.
Healthcare's AI Obsession Is Missing the Point on Nursing Shortages
The op-ed's core argument: nearly 90% of AI investment in hospital operations targets physician-facing tools, clinical decision support, and revenue cycle automation — categories that marginally affect nursing workload. Meanwhile AI tools that directly address nursing's core pain points (patient assignment optimization, float pool scheduling, real-time workload balancing, early deterioration alerts that reduce code responses) remain underfunded and under-adopted. The piece cites internal survey data from three large health systems showing nurses rate documentation burden as their #3 pain point, while health system CIOs rate it #1 — a priority misalignment driving product investment in the wrong direction.
27% of Healthcare Orgs Now Deploying AI Across Multiple Functions — HFMA Survey
The HFMA survey (n=312 healthcare finance executives) found 27% deploying AI at scale across multiple functions, 53% in active pilots, and 20% still in evaluation or not yet started. The 56% who believe operational and technology investment will stabilize their organization's finances represents a significant sentiment shift from 2024, when only 38% held that view. The data is primarily self-reported by finance leadership, which critics note may overstate deployment breadth relative to clinical reality on the ground.