Clinical & Diagnostics
UnitedHealth Group Is Making a $3 Billion Bet on AI — What Does It Mean for Patients?
UHG employs 22,000 software engineers, with over 80% now using AI to write code or build agents. The company is deploying AI across claims auditing, fraud detection, clinical documentation, and billing code selection. It also launched "Avery," a generative AI companion now reaching 6.5 million employer-sponsored members and 160,000 Medicare Advantage members — expanding to 20.5 million total members by year-end.
AI in Cardiovascular Care: From Promise to Practice
The JRC review covers AI applications in ECG interpretation, imaging-based risk stratification, and predictive models for acute cardiac events. Deep learning systems analyzing echocardiograms and CT angiograms now match or exceed specialist accuracy on specific subtasks. The report calls for regulatory harmonization and outcome-linked deployment frameworks to move tools from pilots into standard care.
AI Scribes Cut EHR Charting Time — But Only When Clinicians Use Them Consistently
In the study, clinicians using AI scribes for more than 50% of visits saw twice the reduction in total EHR time and three times the reduction in documentation time compared to lower-frequency users. Only 32% of users reached that usage threshold. The findings suggest deployment strategies need to focus on sustained workflow integration, not just rollout, to realize population-level gains.
UnitedHealthcare Launches Avery, a Generative AI Companion for Members
Avery is a generative AI member-facing tool built on a large language model integrated with UHC's plan data, claims history, and provider networks. Currently live for 6.5 million employer-sponsored and 160,000 Medicare Advantage members, UHC plans to scale to 20.5 million commercial, Medicare, and Medicaid members by end of 2026. The product represents a shift from call-center deflection toward proactive care coordination.
Research & Science
Clinical AI Has Boomed — Stanford-Harvard Report Separates What Works From What Doesn't
The ARISE team synthesized the most impactful clinical AI studies from 2025 across radiology, primary care, and urgent care. Strongest results appeared in early-warning and risk-stratification models trained on large EHR datasets. The report documented over-reliance risks — clinicians following incorrect AI recommendations even when errors were detectable. It calls for real-world validation frameworks that go beyond controlled research settings and match performance claims to actual clinical outcomes.
Deep Learning Bridges Pathology and Radiology in a Single Diagnostic Framework
The model uses convolutional neural networks (CNNs) and Transformer-based attention architectures to jointly process DICOM imaging and digital pathology slide features. Trained and evaluated on oncology datasets, the multimodal approach outperformed unimodal baselines on classification accuracy and disease characterization, particularly for cancers where radiological and histological findings must be reconciled. The system outputs a unified risk score and image-aligned explanations.
AI-Driven Drug–Target Interaction Prediction Charts a Roadmap for Precision Medicine
The review covers deep learning architectures including protein language models, AlphaFold-derived structural representations, and graph neural networks applied to drug-target interaction (DTI) prediction. It addresses key challenges — data sparsity, model interpretability, and generalizability across disease contexts — and proposes a federated learning framework for multisite training without sharing patient-level genomic data. Applications span oncology, rare disease, and CNS drug design.
Merck and Mayo Clinic Launch AI-Enabled Drug Discovery and Precision Medicine Collaboration
The collaboration applies advanced analytics and AI to Mayo Clinic's multimodal clinical dataset — combining imaging, genomic, proteomic, and longitudinal EHR data — to identify novel drug targets and patient stratification signals. Research programs focus on disease areas where precision biomarkers remain elusive. The partnership includes co-development of next-generation sequencing oncology tests and AI-powered patient matching for clinical trials.
Policy & Regulation
Aidoc Secures FDA Clearance for Healthcare's First Comprehensive Foundation Model AI
Aidoc's clearance covers its CARE foundation model, which combines 11 newly cleared indications with 3 previously cleared ones — including pulmonary embolism, intracranial hemorrhage, and aortic dissection — all powered by a single underlying model rather than 14 separate algorithms. The FDA reviewed the CARE architecture holistically, representing a meaningful shift in how multi-task AI systems navigate the regulatory pathway.
FDA Announces Sweeping Rollback of Oversight for AI-Enabled Devices and Wearables
The FDA's revised policy exempts from oversight clinical decision support software that delivers a single recommendation, provided other non-device criteria are met. Products in scope include certain wearable wellness monitors and software that doesn't independently drive treatment decisions. Critics note the boundary between "advisory" and "actionable" AI is blurring in clinical practice, and that the policy creates an unlevel playing field for rigorously validated tools competing against unreviewed products.
Utah Charts a National Model for Regulating Mental Health AI
Utah's framework creates a "safe harbor" for mental health AI agents that implement defined safety guardrails — including mandatory crisis escalation pathways, session data privacy protections, and restrictions on deceptive advertising. The law targets AI chatbots and autonomous therapy assistants deployed directly to consumers, distinguishing them from clinical decision support tools subject to FDA oversight. The framework deliberately avoids mandating specific technical architectures, focusing instead on behavioral and outcome standards.
TEFCA Hits 500 Million Health Records Exchanged as HHS Leverages AI to Reduce Burden
TEFCA's FHIR-based exchange backbone now enables AI systems to access unified longitudinal patient data across disparate EHRs without costly point-to-point integrations. ONC's draft USCDI v7, released January 29, 2026, proposes 29 new data elements including nutrition information exchange and adverse event reporting fields, directly enabling more precise AI model training and clinical decision support. HHS is piloting AI-driven prior authorization streamlining on top of this infrastructure.
Industry & Business
Qualified Health Raises $125M Series B to Scale Enterprise AI Governance at Health Systems
Qualified Health's platform covers the full AI lifecycle: vendor vetting, clinical validation, deployment with built-in clinician oversight, auditability, and continuous post-deployment monitoring. The company now supports 400,000 users representing ~5% of U.S. hospital revenue. The $125M round was led by New Enterprise Associates; Anthropic, Menlo Ventures' Anthology fund, Transformation Capital, and GreatPoint Ventures also participated. Total raised: $155M, valuation between $500M and $1B.
Jimini Health Raises $17M to Launch AI Chatbot for Complex Mental Health Conditions
Sage is an LLM-powered mental health platform designed specifically for complex diagnoses including treatment-resistant depression, bipolar disorder, and schizophrenia — conditions largely underserved by earlier consumer AI tools calibrated on mild-to-moderate presentations. Jimini's go-to-market is B2B2C through behavioral health organizations rather than direct-to-consumer, pairing Sage with clinical oversight infrastructure to manage elevated risk populations.
AI Now Accounts for Nearly Half of All Healthcare Investment, Silicon Valley Bank Reports
SVB's 17th Healthcare Investments and Exits Report recorded more healthcare AI deals over $300M in 2025 than in any prior year, exceeding the sector's overall 2021 peak. AI-enabled healthcare startups in H1 2025 commanded an 83% funding premium over non-AI peers — $34.4M average round size vs. $18.8M. Top funded categories: non-clinical workflow automation (scheduling, billing, prior auth), ambient clinical documentation, and data infrastructure. AstraZeneca's acquisition of Modella AI and Amgen's deal for Dark Blue Therapeutics were among the year's notable M&A moves.
ChatGPT Now Handles 1.6–1.9 Million Health Insurance Questions Per Week
OpenAI's analysis of ChatGPT usage patterns found health insurance navigation (ACA plan selection, prior authorization appeals, EOB explanation, coverage disputes) represents the largest single domain within health-related queries. Users are uploading itemized bills and getting real-time analysis of duplicate charges, improper coding, and potential Medicare violations. The trend highlights a massive unmet need and raises questions about accuracy, liability, and whether AI-generated insurance advice constitutes practice of law or medicine.
Social Buzz
AI in the Mental Health Workforce Is Met With Fear, Pushback — and Genuine Enthusiasm
The piece focuses on ambient documentation AI (similar to Nabla, Abridge, and DAX Copilot) entering behavioral health, where session sensitivity is far higher than in primary care. Core tension: AI that summarizes therapy sessions necessarily processes protected mental health disclosures, raising HIPAA-adjacent privacy concerns beyond standard medical documentation. Clinicians at community mental health centers — which disproportionately serve vulnerable populations — report they lack IT infrastructure to deploy tools safely, widening the technology gap between large and small practices.
Public Trust in Healthcare AI Is Slipping — Down 10 Points Since 2024
The survey found that belief in AI's efficiency in health processes dropped from 64% to 55% since 2024. A majority (53%) hold negative sentiments about AI integration in care, citing dehumanization, data privacy, and algorithmic bias as top concerns. The trust paradox: 88% of current AI health users trust the technology vs. 38% among non-users — suggesting experience drives acceptance but adoption barriers are high. Separately, 51% of adults report having used AI to make an important health decision without consulting a physician.
Nurses Are Being Asked to Use AI Tools They Don't Trust and Were Never Trained On
The Nurse.org survey found 40% of nurses have no input into AI tool selection at their institutions, and the majority report receiving inadequate training before being expected to use AI-assisted scheduling, documentation, or clinical decision support systems. The data comes alongside a separate MedCity News analysis arguing that hospitals are chasing AI-driven efficiency gains while underfunding the structural staffing fixes — ratios, pay, working conditions — that would actually address burnout and retention. AI-only approaches to the nursing shortage are called out as a category error.