🩺Stanford’s AI Predicts Disease Risk From Sleep Data

🩺Stanford’s AI Predicts Disease Risk From Sleep Data

Researchers at Stanford Medicine and collaborating institutions have developed an artificial intelligence model called SleepFM that can predict a person’s risk of developing more than 100 health conditions from a single night’s sleep data. The model was trained on nearly 600,000 hours of polysomnography recordings from 65,000 participants, making this the first AI system to analyze such large-scale sleep data for disease risk prediction.

SleepFM uses complex physiological signals, including brain activity, heart activity, respiratory patterns, leg and eye movements, collected during a standard overnight sleep study to understand patterns in sleep that correlate with future health outcomes.

After training, the model was fine-tuned and tested on long-term health outcomes from a subset of 35,000 patients whose data were paired with their electronic health records, with up to 25 years of follow-up. Across more than 1,000 disease categories, SleepFM identified 130 conditions that could be predicted with reasonable accuracy from sleep data alone, including Parkinson’s disease, dementia, heart attack, breast and prostate cancers, and overall mortality, with several showing concordance indices above 0.8.

Stanford’s researchers emphasize that sleep captures rich physiological information that has historically been underutilized; by treating sleep recordings as a dense multimodal dataset, the AI system effectively learns a “language of sleep” that reveals latent signals associated with disease risk.

What this indicates broadly: If AI models like SleepFM can reliably extract long-term health insights from routine physiological data, they may reshape how early disease risk is assessed and understood. That does not mean these systems are clinical tools yet, nor that they replace human medical judgment, SleepFM’s performance is a research milestone, not a ready-to-deploy diagnostic product. But the study illustrates how AI can repurpose existing medical data in novel ways to reveal latent health indicators, potentially influencing future research priorities, preventative health strategies, and the integration of AI into clinical and research workflows.

📉 Trendlines: JPMorgan Replaces Proxy Advisers With AI

JPMorgan Chase’s asset and wealth management division has replaced external proxy advisory firms with an internal artificial-intelligence system to support shareholder voting decisions, according to Yahoo Finance. The bank will no longer rely on firms such as ISS or Glass Lewis for proxy voting recommendations and instead use its own AI-driven platform, known internally as Proxy IQ.

The article reports that Proxy IQ is designed to analyze voting data from more than 3,000 annual shareholder meetings and present that information directly to JPMorgan’s portfolio managers. JPMorgan said the system will provide research and analysis that was previously supplied by proxy advisers, allowing the firm to handle the process internally.

Proxy advisory firms have traditionally provided institutional investors with research and recommendations on corporate governance matters, including board elections, executive compensation, and shareholder proposals. JPMorgan’s move makes it the first major investment manager to fully replace those services with an in-house AI system, according to the article.

JPMorgan executives cited concerns about errors and inconsistencies in proxy advisers’ recommendations as part of the rationale for the shift. The bank emphasized that portfolio managers will still make the final voting decisions, using the AI system as an analytical tool rather than an automatic decision-maker.

While the article does not discuss labor markets, employment, or broader economic effects, it does document a clear substitution: a function historically performed by external advisory firms is now being handled by an internal AI-based system. That fact alone makes the move notable.

In light of recent public warnings from leading AI researchers that 2026 could mark the beginning of a “jobless boom,” JPMorgan’s decision can be read as a concrete example of how institutions are beginning to restructure professional workflows around AI systems. The article does not claim this will reduce jobs, but it does show how AI is already being used to replace entire categories of outsourced specialist services within large organizations.

The significance of the move is therefore narrow but clear: AI is no longer confined to experimentation or support roles, it is being trusted with core institutional processes that were previously handled by human-led external advisers.

🛍️Google Unveils Universal Commerce Protocol to Power AI-Driven Shopping

Google announced a new open standard called the Universal Commerce Protocol (UCP) for AI agent-based shopping at the National Retail Federation’s annual conference in New York. The protocol is designed to let AI agents work across multiple stages of the customer buying process, from product discovery through checkout and post-purchase support, without requiring bespoke connections between individual systems.

UCP was developed in collaboration with major retailers and platforms including Shopify, Etsy, Wayfair, Target, and Walmart, and is interoperable with existing agent-focused protocols such as Agent Payments Protocol (AP2), Agent2Agent (A2A), and Model Context Protocol (MCP).

Google said it will soon use UCP to power a new shopping experience on eligible product listings in AI Mode in Search and the Gemini apps, enabling users to complete purchases with Google Pay and shipping information stored in Google Wallet, with PayPal support coming soon.

A notable feature Google highlighted is that merchants can offer special discounts directly within AI Mode recommendations, potentially aligning product discovery and purchase incentives more closely for consumers.

The announcement also noted that many merchants are already exploring branded AI agents embedded directly in search to answer questions and assist shoppers.

In short, Google’s Universal Commerce Protocol is an open standard aimed at simplifying and standardizing AI-mediated commerce across platforms, making it easier for AI assistants to interact with retailers and payment systems to help users discover, evaluate, and buy products seamlessly within conversational interfaces.

🧩 Closing Thought

Taken together, these stories illustrate an AI landscape shifting from augmentation to integration. Stanford’s SleepFM shows how AI can extract long-term health signals from existing medical data, JPMorgan demonstrates AI being trusted with analytical functions at the core of institutional governance, and Google’s commerce protocol points toward a future where AI agents act as intermediaries in economic transactions. What unites these developments is normalization rather than novelty, AI is being woven into systems that already exist, making the change less visible but potentially more durable. The next phase of adoption will be defined less by what AI can do and more by where institutions are willing to allow it to participate, advise, and act.