AI Sector Daily Digest — May 31, 2026

Five stories from the past 24 hours.

At its annual I/O developer conference, Google announced a redesigned search interface — an "intelligent search box" built for long, conversational queries — alongside generative UI that can produce interactive widgets and visualizations rather than lists of links. Background "information agents" can now track topics continuously and alert users when conditions change; a mini-app builder called Antigravity lets users assemble small personalized tools inside Search using natural language. AI Overviews reach 2.5 billion monthly users; AI Mode has passed 1 billion.

On the model side, Google released Gemini 3.5 Flash — described as faster and cheaper than competing models — as the engine powering search and Google Workspace. Gemini Omni, a "world model" that generates video from any input type (text, image, audio), is rolling out to Google AI Plus, Pro, and Ultra subscribers globally. Gemini Spark, a new agentic tool that can take real-world actions such as booking flights and adding items to Instacart carts, is live now for AI Ultra subscribers ($100/month) in the US.

Speaking at the Choose France investment summit, SoftBank Group chairman Masayoshi Son said the company will spend up to €75 billion (roughly $87 billion) to develop 5 gigawatts of AI data center capacity across France — its largest European infrastructure bet. The first phase targets 3.1 GW across three sites in the Hauts-de-France region (Dunkirk, Bosquel, Bouchain) by 2031. A joint venture with French AI infrastructure firm Sesterce has already been selected to build the 1 GW Bosquel campus, which will create 400 permanent jobs and comes with a €10 million endowment for local schools and businesses. French economic minister Roland Lescure called the commitment "a testament to President Macron's ambition to position France as a leading destination all along the AI value chain."

London-based Inherent came out of stealth Wednesday with a $50 million seed round co-led by Index Ventures and Radical Ventures. Nvidia's venture arm NVentures also participated. The founding team — Tantum Collins, Edward Hughes, Louis Kirsch, and Kaloyan Aleksiev — met at DeepMind, where Collins and Hughes worked on cooperative AI research. Collins previously worked on AI policy at the Biden White House; Matt Clifford, former UK government AI tsar, joined as an adviser.

The company's platform, Faraday, pairs human researchers with self-improving AI agents to identify which scientific questions are worth asking rather than just answering known ones. Inherent is structured as a public benefit corporation, an unusual choice for a venture-backed lab. Index partner Danny Rimer described the thesis this way: "Most AI is built to answer questions. What it can't do yet is figure out which questions are worth asking."

OpenAI launched Rosalind Biodefense on May 29, making its GPT-Rosalind life-sciences reasoning model available at no cost to vetted developers, academic institutions, nonprofits, and US government and allied-nation agencies working on pandemic preparedness and biodefense. OpenAI covers all access costs for approved participants.

GPT-Rosalind was first introduced in April 2026 for drug discovery and protein reasoning. OpenAI's internal benchmarks put it ahead of GPT-5, GPT-5.2, and GPT-5.4 on chemistry, biochemistry, and experimental design tasks. Early partners include Lawrence Livermore National Laboratory (combining the model with supercomputing for medical countermeasure design), Johns Hopkins Applied Physics Laboratory (protein engineering for therapeutic screening), and CEPI (rapid vaccine development for emerging threats including Bundibugyo ebolavirus). Applications go through two tracks: a developer path open globally, and a government path limited to agencies with explicit public-health or biodefense mandates.

Microsoft is moving GitHub Copilot from a flat subscription rate to a token-usage model effective June 1. Developers will be charged based on how many tokens they consume, not a fixed monthly fee. The shift has sparked sharp criticism on Reddit and X, with some users reporting that their costs will jump from around $29/month to over $750, and one sharing a screenshot appearing to show a billing increase from $50 to roughly $3,000.

Not everyone objects: some developers argue that token-heavy costs are driven by "vibe coding" — high-iteration sessions with little architectural discipline — and that careful users will see little change. Critics, though, point out that Microsoft actively designed and marketed the flat-rate model to encourage heavy usage, and is now penalizing users who relied on those incentives. Microsoft did not respond to TechCrunch before publication.