AI Digest · Archived

AI Digest

An archived edition. For the current digest, see the latest.

Edition of June 3, 2026 · 42 sources · 764 articles · 48h window ← Latest digest · All past digests Subscribe via RSS

  1. Microsoft Build 2026 goes all-in on agents and its own models.

    Microsoft unveiled seven in-house MAI models, led by the 1-trillion-parameter MAI-Thinking-1 reasoning model and a Copilot-tuned MAI-Code-1-Flash, loosening its dependence on OpenAI. It also launched Scout, an always-on personal assistant inside Microsoft 365, and Project Solara, an Android variant built for AI agents instead of apps.

    Sources The Verge ·Simon Willison ·Ars Technica ·TechCrunch

  2. Trump signs a watered-down AI executive order.

    After industry pushback, the president scrapped a tougher draft and signed an order creating only a voluntary framework for AI labs to share frontier models with the government before release. Policy experts warn the "trusted partner" access could still let officials pick winners and losers.

    Sources The Verge ·TechCrunch ·The Register ·Politico

  3. OpenAI pushes Codex from coding tool to everyday work platform.

    OpenAI rolled out role-specific Codex plugins for analysts, marketers, designers, investors and bankers, plus a "Sites" web-hosting feature and inline "Annotations," reframing Codex as an operating environment for knowledge work. In parallel, GPT-5.5, GPT-5.4 and Codex became generally available on Amazon Bedrock.

    Sources OpenAI ·VentureBeat ·TechCrunch ·AWS

  4. Anthropic files for what could be the largest IPO ever.

    The maker of Claude confidentially submitted IPO paperwork to the SEC, weeks after topping OpenAI's valuation and just after SpaceX's own IPO announcement. The milestone was punctuated by a Claude outage on the very day of the float news.

    Sources Wired ·The Register

  5. The cost of running AI starts to bite.

    GitHub Copilot's switch to usage-based pricing triggered a revolt, with some developers burning a month's credits in a single day and threatening to leave, while Uber capped employee AI spending after blowing its budget in four months. Vendors are responding with local-first hardware like Microsoft's Surface RTX Spark Dev Box and Perplexity's hybrid local-cloud inference to dodge per-token cloud costs.

    Sources Ars Technica ·The Register ·TechCrunch ·VentureBeat

  6. AI agents become the new attack surface.

    Hackers simply asked Meta's AI support chatbot to hand over high-profile Instagram accounts, and it worked, while researchers disclosed "BadHost," a Starlette auth-bypass flaw exposing AI agent infrastructure behind 325 million weekly downloads. Microsoft answered with MXC, an OS-level sandbox for agents already backed by OpenAI and Nvidia.

    Sources Ars Technica ·Simon Willison ·InfoQ ·VentureBeat