ma2tic

The brief · 2 June 2026

The AI brief, 2026-06-02

Seven Microsoft models, a Trump executive order finally signed, and an AI tracing back to the origins of life: the day mixes scales. Between banks still hunting for their ROI and the WHO setting its guardrails, one common thread: measure before you believe.

3 min read J / K to navigate

A Generative Model Predicts Your Biological Age From Seven Data Types

AURORA combines genome, microbiome, proteome, and other biological layers across more than 425,000 people to estimate your true metabolic age and simulate the effect of an intervention before you actually try it. This is still research, not something you'll be using in ChatGPT tomorrow. But it's the kind of building block that will likely show up in lab tests or health-tracking apps within two or three years.

practitioners › Published in Cell Metabolism on June 2, 2026, covering more than 425,000 individuals and integrating seven omic and phenotypic modalities.

Cell Metabolism →

WHO Weighs In on AI's Role in Health Policy

No blanket ban, no green light either. The WHO maps out where AI can genuinely help shape health policy decisions, and where it can backfire. If you're using ChatGPT to make sense of a study or unpack a health recommendation, this is still a distant signal for now. But it's laying the groundwork for firmer rules down the line.

practitioners › Discussion paper published June 2, 2026, no binding recommendations at this stage.

WHO →

Corporate Insight surveyed 2,000 people on trust in banking AI

The report cross-references 21 financial sectors with consumer research to measure a real gap: banks are rolling out AI faster than customers are willing to accept it. If you work in finance or just use a banking app with an AI assistant, this explains why certain features stay opt-in rather than on by default.

practitioners › Study published June 2, 2026, based on 2,000 respondents and consumer interviews across 21 financial verticals.

corporateinsight.com →

Microsoft drops seven in-house models in one go

Microsoft is done relying solely on OpenAI and is building out its own model lineup, covering everything from reasoning to image generation to voice transcription. That means more options if you're using Copilot or Azure tools, and a clear signal that Microsoft wants technical independence.

practitioners › MAI-Thinking-1 (reasoning), MAI-Code-1-Flash (coding) and MAI-Image-2.5 are already available via OpenRouter, plus a customization option called Frontier Tuning.

Microsoft AI →

Trump Signs His AI Executive Order, Betting Everything on Voluntary Compliance

No mandatory licensing, no new agency, no heavy constraints for US labs. Washington is choosing collaboration with the private sector over regulation, in sharp contrast to California and Illinois, which just passed their own laws. The gap between the federal government and the states is becoming hard to ignore.

practitioners › Executive Order 14409, signed June 2, 2026: voluntary frameworks for frontier models, cybersecurity as top priority, targeted prosecution of AI-assisted crimes.

Maison Blanche →