AP and Lateral

Every doctor in Abu Dhabi now has an AI assistant inside the shared patient record

AP: The Department of Health and M42's Abu Dhabi Health Data Services have added an AI assistant to Malaffi, the emirate's health information exchange. It summarises a patient's history and answers a doctor's questions about it. Malaffi holds data on more than 14 million patients, and the rollout covers the whole emirate, not a pilot group. M42 launched its own clinical AI platform, Ossi, with Microsoft in October. The announcement does not say which model powers the Malaffi assistant.

Lateral: this is a language model working across a population-scale record, used at the point of care. A summary that leaves out an allergy or an anticoagulant is a safety event, so the evaluation matters more than the launch.

Not disclosed: the model behind it, accuracy or error rates, time saved, uptake, and when M42's promised evaluation with clinicians will report.

Sources: tbreak, DH Arab, The National on Ossi

Abu Dhabi launches an AI Clinician Framework and a Trusted Research Environment

AP: Both were launched by the Department of Health at the Abu Dhabi Future Health Summit, according to its round-up published on 11 November. The summit drew more than 17,300 visitors from 85 countries. The framework is meant to guide research and prepare the ground for regulating AI that practises with autonomy. The research environment allows work on health and genomic data inside a secure setting. The department also signed with Isomorphic Labs on AI drug discovery.

Lateral: a regulator is writing rules for autonomous clinical AI before it arrives. Anyone planning a clinical AI study in the Gulf should read the framework first.

Source: Abu Dhabi Media Office

Saudi National Guard hospitals take Lunit's chest X-ray and mammography AI

AP: Lunit will supply INSIGHT CXR and INSIGHT MMG to the Ministry of National Guard Health Affairs. solutions by stc is the systems integrator, and results will appear inside the existing PACS. Lunit signed with Dr. Sulaiman Al Habib Medical Group in 2023, so this moves it from private providers into a government system.

Lateral: integration into PACS is what decides whether radiologists use a tool. The useful follow-up is local performance data on Saudi patients.

Not disclosed: contract value, number of licences and the deployment schedule.

Sources: Lunit release, Korea Biomedical Review

ART Fertility Clinics and Conceivable to explore robotic, AI-assisted IVF

AP: The two signed a memorandum of understanding at the summit. It covers laboratory infrastructure, workflow, staff training, and the validation and regulatory steps needed before any deployment.

Lateral: it is an agreement to explore, not a service. Nothing is treating patients yet.

Source: Healthcare MEA

Review: digital pathology and AI in Saudi Arabia held back by cost, workforce and data

AP: An open-access review published on 4 November maps what stands between Saudi laboratories and routine AI-assisted histopathology. It is a synthesis of existing literature and reports no new experiments.

Lateral: a review cannot show whether these tools work on Saudi slides. Read it as a map of the obstacles, not as evidence of performance.

Sources: the paper (DOI), Bioengineer.org summary

Models know more Arabic medicine than they can say

AP: A team from New York University Abu Dhabi and Cleveland Clinic Abu Dhabi finds that Arabic medical knowledge sits in a model's middle layers but often fails to reach its answer. Adapting only the lower layers beat adapting the whole network on medical multiple-choice questions.

Lateral: poor Arabic performance may be fixable cheaply, without retraining from scratch.

Source: arXiv 2608.00207 (preprint, not peer reviewed)

MedAraBench: 24,883 Arabic medical exam questions across 19 specialties

AP: Presented at ICLR 2026, built by digitising teaching material written by medical professionals in the region. Every model tested lost accuracy as questions demanded more clinical reasoning. The dataset and scoring scripts are public.

Lateral: exam questions test recall and reasoning on paper. A high score does not show a model is safe with Arabic-speaking patients.

Sources: arXiv 2602.01714, ICLR 2026 listing