← AI Intelligence Hub

Clinical & Diagnostics

Butterfly Network Wins First-Ever FDA Clearance for Blind-Sweep AI Gestational Age Ultrasound Tool

Executive Brief Any healthcare worker — no sonography training required — can now estimate a pregnant patient's gestational age in under two minutes using Butterfly's handheld ultrasound and a new AI tool that does all the image interpretation automatically. The clearance opens maternal care to clinics and communities that have never had a sonographer on-site.

Butterfly's Gestational Age Tool received FDA 510(k) clearance on March 30, 2026, becoming the first cleared blind-sweep ultrasound AI in the U.S. The model was trained on more than 21 million images across diverse demographics and delivers biometry-equivalent gestational age assessments for patients between 16 and 37 weeks — matching sonographer accuracy. The three-step workflow (enter fundal height, apply gel, perform guided sweeps) requires no screen interpretation by the user. Backed by the Bill & Melinda Gates Foundation, the tool is already deployed in Malawi and Uganda; U.S. clearance now enables rural expansion at scale.

Aidoc Secures FDA Clearance for Healthcare's First Comprehensive Foundation Model AI — Triaging 14 Acute Conditions from One Model

Executive Brief For the first time, a single FDA-cleared AI model can triage 14 different life-threatening conditions on a CT scan — eliminating the fragmented, condition-by-condition AI stack hospitals have wrestled with for years. Emergency departments get one workflow, one integration, and far fewer false alarms.

Aidoc's CARE™ foundation model received FDA clearance covering 11 newly cleared acute indications plus three previously cleared ones, all running on a single abdominal CT AI architecture. The pivotal study reported mean sensitivity of 97% and mean specificity of 98% across the 11 new indications — with performance up to 98.5% sensitivity and 99.7% specificity on individual findings. The system delivers roughly an order-of-magnitude reduction in false alerts compared to best-in-class single-condition solutions, and the CARE roadmap includes planned expansion to all CT and X-ray workflows within 18 months.

Athenahealth Rolls Out AI-Native EHR Platform and Launches Ambient Scribe for 100,000+ Providers

Executive Brief Athenahealth is giving its entire ambulatory network — more than 100,000 provider customers — an AI-native upgrade to its core EHR platform at no extra cost. The centerpiece is athenaAmbient, an ambient documentation scribe that converts patient-physician conversations directly into structured clinical notes, eliminating the most time-consuming part of a provider's day.

The AI-native athenaOne platform refresh includes athenaAmbient (in testing since February 2026), AI-enhanced document services, intelligent clinical summaries, and improved interoperability tooling. Athenahealth also became the first healthcare IT company to implement TEFCA at scale across all eligible providers in its 160,000-strong network. The company is piloting a Model Context Protocol (MCP) server for athenaOne — claimed as a first for the industry — enabling AI models to communicate directly with EHR data without custom integration work.

Research & Science

Merck and Mayo Clinic Form Landmark AI Drug Discovery Partnership, Opening Access to Genomic and Multimodal Clinical Data

Executive Brief Merck now has direct, secure access to Mayo Clinic's de-identified clinical and genomic data — including imaging — to train AI models that identify drug targets earlier and with greater precision. The collaboration, Mayo's first of this scale with a global pharma company, targets inflammatory bowel disease, atopic dermatitis, and multiple sclerosis first.

The partnership gives Merck access to Mayo Clinic Platform Orchestrate, a secure multimodal repository spanning genomic sequences, clinical records, imaging, and biorepositories. Merck will deploy AI-enabled virtual cell technologies to model disease biology, improve target identification, and accelerate early-stage development decisions. The collaboration integrates Merck's ML research capabilities with Mayo's Platform architecture, targeting three high-need therapeutic areas initially, with the intent to generate in silico biological insights before wet-lab validation begins.

Stanford-Harvard ARISE Report: Clinical AI Has Boomed — But Top Models Still Make Severe Errors in 12–15% of Cases

Executive Brief The first-ever State of Clinical AI Report — produced by a joint Stanford-Harvard research network — finds that AI is already embedded in everyday care, but honest measurement reveals that even leading models commit severe clinical errors more than 1 in 10 times. The report is the field's most comprehensive attempt to separate clinical AI that actually helps from AI that merely looks impressive in a demo.

The ARISE (AI Research and Science Evaluation) network synthesized the most influential clinical AI studies published in 2025 across Stanford, Harvard, and affiliated health systems. Key quantitative finding: top-performing AI models made between 12 and 15 severe errors per 100 clinical cases; worst-performing systems exceeded 40 severe errors per 100 cases. The report identifies brittleness in uncertainty estimation — models fail to recognize when they don't know — as the most persistent systemic risk in current clinical deployment. Full report available at arise-ai.org/report.

Nature Study: Deep Learning Fuses Pathology and Radiology Into a Single AI-Assisted Diagnostic System

Executive Brief A new AI framework combines microscopic pathology images and macroscopic radiology scans into one unified diagnostic model — a capability that has historically required separate specialists working in silos. The result is more accurate diagnoses from a single AI system, with implications for oncology, rare disease, and settings where subspecialty access is limited.

The paper presents a deep learning image classification framework that integrates pathology's cellular morphology analysis with radiology's anatomical and functional imaging into a multimodal pipeline. The system uses cross-modal attention mechanisms to combine spatial features from histological slides and volumetric scan data, with validation across oncology use cases demonstrating statistically significant improvement in diagnostic accuracy versus single-modality baselines. The work addresses a key bottleneck in clinical AI: most existing tools operate on one data type in isolation, missing complementary diagnostic signals.

Policy & Regulation

WHO Calls Generative AI a Public Mental Health Concern, Releases Responsible AI Framework for Mental Health Tools

Executive Brief The World Health Organization formally classified generative AI's impact on mental health as a public health concern — placing it alongside other regulated health risks — and issued guidance calling on governments and health systems to act now, before harms compound. The move signals a coming wave of national-level regulation on AI therapy apps and mental health chatbots.

The guidance emerged from a January 29, 2026 expert workshop convening over 30 international specialists in AI, mental health, ethics, and public policy. Three core recommendations: (1) classify GenAI's mental health impact as a public health concern requiring government response; (2) integrate mental health impact assessments into AI product monitoring, with specific attention to emotional dependence and long-term outcomes; (3) mandate co-design of mental health AI tools with clinical experts and people with lived experience. WHO is simultaneously establishing a global Consortium of Collaborating Centres on AI for Health spanning all six WHO regions.

TEFCA Hits 500 Million Health Records Exchanged — A 4,900% Jump in 12 Months

Executive Brief America's national health data exchange network crossed 500 million records shared — up from just 10 million in January 2025. That acceleration means AI systems trained and deployed on TEFCA-connected data now have access to a fast-growing, nationwide, real-world clinical dataset, fundamentally changing what population-level AI models can learn and do.

The Trusted Exchange Framework and Common Agreement (TEFCA) — a government-backed, FHIR-enabled interoperability initiative — processed nearly 500 million record exchanges, representing a 4,900% increase since January 2025. HHS projects over $19.2 billion in administrative cost savings over the next decade from health IT regulations enabling electronic prior authorizations. Simultaneously, ASTP/ONC released the draft USCDI v7 proposing 29 new standardized data elements, and $20+ million was committed to nine Behavioral Health IT pilot programs integrating SAMHSA data into TEFCA-compatible workflows.

FDA Cuts Red Tape on Clinical Decision Support Software, Allowing Single-Recommendation AI Tools to Bypass Review

Executive Brief The FDA softened its oversight of clinical decision support software: AI tools that provide a single clinical recommendation — without requiring a clinician to independently review the underlying logic — can now reach the market without FDA premarket review, as long as they meet existing exemption criteria. The change accelerates AI deployment into clinical settings but raises new questions about safety accountability.

The FDA's revised CDS guidance removes the requirement that single-recommendation software — previously considered "non-device" only when clinicians could independently review and understand the basis of recommendations — must undergo 510(k) clearance. This follows CPT 2026's inclusion of 288 new codes covering digital health and AI services, and CMS expansion of payment policies for digital mental health treatment devices. The policy shift is concurrent with an overall FDA posture toward lighter-touch regulation of wearables and AI-enabled devices announced in January 2026.

Industry & Business

Jimini Health Raises $17M for Sage — A Clinician-Supervised AI Behavioral Health Chatbot

Executive Brief Jimini Health's Sage chatbot keeps behavioral health patients engaged between therapy sessions — contacting them with reminders, check-ins, and structured support — while the human clinical team monitors every conversation and retains all care decisions. The $17M seed round positions Sage as the responsible alternative to fully autonomous AI therapy apps that have drawn regulatory scrutiny.

The $17M round was led by M13, with participation from Town Hall Ventures, LionBird, Zetta Venture Partners, and OneMind, bringing Jimini's total capital to over $25M. Sage is trained to follow individualized care plans written by human clinicians — it does not improvise responses — and clinicians retain full access to all patient-Sage conversation logs, with automated alerts triggering when the system detects deterioration. Jimini plans to use the funding to scale EHR integrations and expand partnerships with behavioral health organizations.

Qualified Health Raises $125M Series B to Operationalize Enterprise AI Across 500,000+ Health System Users

Executive Brief Qualified Health — which helps hospitals safely adopt and govern AI at scale — just closed a $125M Series B with backing from NEA and Anthropic's investment arm. With 500,000+ users across health systems representing roughly 7% of U.S. hospital revenue, the company is now one of the best-funded enterprise AI platforms in healthcare.

The Series B was led by NEA, with participation from Transformation Capital, GreatPoint Ventures, Cathay Innovation, and Menlo Ventures' Anthology Fund (an Anthropic-partnership AI fund). Qualified Health's platform provides enterprise AI governance infrastructure — safety guardrails, compliance tooling, and deployment pipelines — designed for health systems that need to deploy and manage multiple AI models simultaneously. Current customers include Mercy, Emory Healthcare, University of Rochester Medicine, Jefferson Health, and all eight University of Texas System health institutions. Funds will be directed to product engineering and expanded clinical AI capabilities.

Doctronic Raises $40M After Becoming First AI Platform Legally Authorized to Renew Prescriptions in the U.S.

Executive Brief Doctronic crossed a milestone no AI company has crossed before: Utah legally authorized its AI system to renew prescriptions for patients with chronic conditions — without a human physician reviewing each refill. The $40M Series B follows 15x revenue growth in six months and signals that autonomous AI medicine is moving from concept to regulated practice.

Doctronic's Series B was led by Abstract and Lightspeed Venture Partners, bringing total capital raised to $65M across three rounds in under a year. The platform's core architecture pairs an AI chatbot for symptom intake with telehealth physician handoffs for diagnosis — handling 15 million+ medical conversations with 1 million+ users, and serving 300,000 unique weekly visitors. Since its $20M Series A in September 2025, the company reported 15x revenue growth to 8-figure annualized revenue. Utah's January 2026 autonomous prescription renewal partnership represents the first state-level regulatory authorization for AI prescribing in U.S. history. Expansion into pediatrics is the stated next priority.

J.P. Morgan: AI Now Drives 75% of All Health Tech Deals, Commanding an 83% Funding Premium Over Non-AI Startups

Executive Brief Three of every four health technology investment deals now involve an AI-enabled company, and those AI startups raise 83% more per round than their non-AI counterparts. The J.P. Morgan data makes it official: healthcare AI isn't a sub-segment of health tech anymore — it is health tech.

J.P. Morgan's 2026 health technology report found that AI-focused deals now constitute 75% of all health tech transactions, with AI-enabled healthcare startups capturing 62% of all digital health venture funding in the U.S. in the first half of 2025. AI startups raised an average of $34.4M per round — an 83% premium over non-AI digital health rounds. Series B rounds represent 60% of AI-related transaction volume, and the global healthcare AI market is projected to reach $45.2 billion by end of 2026.

Social Buzz

Viral Op-Ed: "Healthcare's AI Obsession Is Missing the Point on Nursing Shortages" — MedCity News

Executive Brief A widely-shared MedCity News opinion piece argues that hospital systems are deploying AI tools to paper over a nursing shortage that AI cannot solve — and that the real fix requires workforce investment, not algorithmic substitution. The piece sparked a heated LinkedIn and X debate between health system executives and nursing advocates about whether AI is a solution or a distraction.

The op-ed lands in a context where the U.S. faces a deficit of over 250,000 registered nurses, and more than 65% of hospitals have run below full capacity due to staffing shortages. Critics of the AI-first approach note that while tools like predictive burnout modeling, virtual nursing, and robotics (e.g., Nurabot from Foxconn, Moxi from Diligent Robotics) reduce non-medical workload, they do not address the systemic drivers of nurse attrition: pay, conditions, and staffing ratios. The debate has intensified as vendor claims about AI "solving" staffing have multiplied.

Patients Are Using ChatGPT to Decode Medical Bills, Catch Overcharges, and Win Insurance Appeals

Executive Brief A wave of viral patient stories is documenting how people are uploading itemized hospital bills to ChatGPT and getting it to find duplicate charges, incorrect billing codes, and violations of Medicare rules — errors that previously required a medical billing advocate to catch. For the first time, AI is giving ordinary patients parity with the billing systems stacked against them.

OpenAI's ChatGPT has added health insurance navigation features targeting ACA enrollment, coverage explanations, and billing dispute support. Users share itemized bills as structured PDFs or images, and the model identifies anomalies by cross-referencing CPT codes, payer policies, and Medicare fee schedules embedded in its training data. The workflow has evolved from curiosity to a repeatable patient strategy: upload bill → flag suspicious line items → draft appeal letter → submit. Healthcare policy observers note that widespread AI-assisted appeals could create systemic pressure on payer billing practices.

STAT: "Who'll Pay for AI in Healthcare?" Is the Question Nobody Has Answered Yet

Executive Brief STAT's widely-read 2026 outlook piece crystallizes the industry's central unresolved tension: AI in healthcare is proliferating rapidly, but payment models, reimbursement codes, and the business case for payers are still largely missing. Without a sustainable funding pathway, many of today's AI deployments are experiments burning through grant money and VC subsidy — not durable healthcare infrastructure.

The piece identifies three structural payment trends to watch in 2026: (1) CMS CPT code expansion — 288 new codes now cover AI-assisted services, but reimbursement rates remain too low for most vendors to build sustainable businesses; (2) payer cost-sharing models, where AI that demonstrably reduces total cost of care might be paid for by insurers rather than providers; (3) direct-to-employer AI health benefits, where self-insured employers pay for AI tools as part of benefit packages. The reimbursement gap has grown into healthcare AI's biggest structural headwind, with venture investors increasingly scrutinizing "who pays" before funding Series A rounds.