Google's Gemini 3.1 Ultra lands with a 2-million-token context window, Microsoft debuts its own code AI, and governments on three continents tighten AI rules: all in 24 hours.
Google's Gemini 3.1 Ultra is here: 2 million tokens, no transcription needed
Google launched Gemini 3.1 Ultra, calling it its most significant model release of the year. It handles a 2-million-token context window (roughly 1,500 books worth of text), and works natively across text, images, audio, and video without needing to transcribe or convert anything first. It also ships with a Code Execution tool that lets the model write, run, and test code during a conversation, not just generate it.
Why it matters: The gap between what AI can hold in memory and what a human expert can review in a sitting is now enormous. For researchers, analysts, and anyone working with large documents or long recordings, this model is a step change.
Microsoft launches its own AI models: MAI-Code-1-Flash and MAI-Thinking-1
At its Build developer conference, Microsoft announced two models built in-house. MAI-Code-1-Flash converts plain-English descriptions into working application code. MAI-Thinking-1 is a dedicated reasoning model for complex, multi-step problems. Both are positioned to reduce Microsoft's dependence on OpenAI for its developer and enterprise products.
Why it matters: Every major tech company now has its own AI model strategy. For developers and businesses, this means more choices, and the beginning of real competition on price and capability between platforms that until now had a near-monopoly.
Trump signs AI Executive Order: pro-innovation, but with a cybersecurity catch
The White House signed an Executive Order titled "Promoting Advanced Artificial Intelligence Innovation and Security." It removes bureaucratic constraints on AI developers and signals the US will lead through industry freedom rather than heavy regulation. But it also directs the Attorney General to enforce criminal law against anyone who uses AI to illegally access computer systems, making AI-enabled hacking an explicit federal priority.
Why it matters: The US is trying to have it both ways, light on industry, hard on misuse. How these two parts of the order interact in practice will shape what AI companies can and cannot build in the next few years.
EU bans AI deepfakes of real people. UK regulators warn on frontier AI cyber risk.
The EU AI Act now includes an explicit ban on AI systems that generate non-consensual intimate content or child sexual abuse material, moving from guideline to legal prohibition. In the UK, the Bank of England, the FCA, and HM Treasury issued a rare joint statement warning that frontier AI models pose new categories of cyber risk to financial systems. Colorado's comprehensive AI Act takes effect June 30, becoming the first US state law of its kind to go live.
Why it matters: Regulation is no longer hypothetical. Three different governments moved from discussion to enforcement this week. If you build, deploy, or use AI in a professional context, the legal landscape is changing fast.
Cognizant's Project Leap: AI reskilling and job cuts arriving together
Cognizant rolled out Project Leap, a company-wide AI transformation program that combines workforce reskilling with job cuts, reflecting the broader pattern across India's IT sector as AI reduces the need for volume hiring. Separately, India's weight in the MSCI Emerging Markets index fell to a 6-year low of 10.87%, as AI-driven rallies in Taiwan, South Korea, and China push Indian stocks down the rankings. India's deeptech startup base stands at over 3,600 companies, but the jobs transition question is not resolved.
Why it matters: India's IT sector built its global reputation on scale and workforce numbers. AI is restructuring that model faster than reskilling programmes can respond. If you work in Indian IT, the window to build skills is narrower than the headlines suggest.
AI literacy is now law in US schools, and most college students are already using it daily
Maryland's AI Ready Schools Act went live on June 1, requiring the state education department to issue AI guidelines and embed AI literacy into K-12 computer science and workforce standards by June 2027. The State University of New York system has made AI literacy mandatory training for all incoming undergraduates starting Fall 2026. A new survey finds 64% of college students already use AI daily or weekly for coursework, and 61% of elementary educators say their students struggle to distinguish AI-generated content from human-written work.
Why it matters: Schools are being asked to teach AI literacy while students are already ahead of the curriculum. The gap between what students actually know and what they are formally taught is the defining education challenge of this moment.
Colorado's AI Act goes live June 30, the US's first real AI compliance deadline
Colorado's AI Act takes effect in 19 days. It requires companies deploying AI in high-risk settings to run security risk assessments, conduct impact evaluations, and actively prevent algorithmic discrimination. No other US state has a law this specific that is actually in force. Every other state, and the federal government, is watching whether enforcement works, and whether it triggers a race to move AI operations elsewhere. This is the first real test of whether US AI regulation can land without being watered down.
OpenAI opens an AI economics research exchange + NAVER and NVIDIA build sovereign AI at scale
OpenAI launched the OpenAI Economic Research Exchange, a platform for external researchers to study AI's real-world impact on jobs, productivity, and growth, a direct response to growing calls for independent data on AI's economic effects. Separately, NAVER and NVIDIA announced a major sovereign AI infrastructure deal: NAVER will build a data centre starting at 55 megawatts and scaling toward gigawatt capacity in South Korea, running its next-generation HyperCLOVA X models on NVIDIA's DSX platform.