Microsoft Build went all-in on autonomous AI agents, Trump signed a surprise safety order, and Anthropic is now within touching distance of a trillion-dollar valuation.
Microsoft Build 2026: AI agents that watch your inbox and run long tasks alone
Microsoft's annual developer conference ended with a cascade of AI announcements. The headline: "Autopilots": long-running autonomous agents that monitor your inbox, Teams messages, and calendar and act on your behalf without being asked. Also announced: a GitHub Copilot desktop app, the Microsoft Agent Framework (merging AutoGen and Semantic Kernel into one SDK), and the Surface RTX Spark Dev Box, a hardware workstation that can run 120-billion-parameter AI models entirely offline.
Why it matters: The shift from "AI that answers questions" to "AI that handles tasks while you sleep" is no longer a concept, it shipped this week. Developers now have a commercially supported toolkit to build it, and consumers will start seeing these agents inside everyday Microsoft apps soon.
Trump signs AI safety order: big labs must let the government test models before release
President Trump signed an executive order on June 2 directing AI companies to voluntarily submit their most powerful models for government review up to 30 days before public release. The order asks OpenAI, Anthropic, Google, and others to share early access so federal agencies can run security and capability tests. It also creates an "AI cybersecurity clearinghouse" to track vulnerabilities. Participation is voluntary: the order explicitly bars a mandatory licensing regime.
Why it matters: This is the US government's most direct intervention in AI development to date, even if companies can still say no. It signals that Washington is no longer comfortable with being surprised by the next powerful model, and that the conversation about AI oversight has moved from Congress to the White House.
Google's Gemini 3.5 Pro goes generally available: the fastest frontier model yet
Google's Gemini 3.5 Flash, launched at Google I/O in May, is now fully available to everyone, through the Gemini app, Google Search, and developer APIs. The Pro tier is rolling out this month. Gemini 3.5 Flash is four times faster than comparable frontier models, leads on coding and agent benchmarks, and is priced at $1.50 per million input tokens: cheaper than its competitors at similar performance levels.
Why it matters: Speed and cost are the two things that determine whether AI tools get built and actually used. Gemini 3.5 Flash being both frontier-grade and fast is a meaningful upgrade for developers building products, and for everyday users who will notice AI responses getting snappier inside Search and apps.
India is now the world's third-largest AI startup hub, and the government is adding fuel
India now has over 4,500 active AI startups, ranking third globally behind only the US and China. Indian AI startups raised $1.48 billion in Q1 2026 alone, with AI accounting for 38% of all startup funding that quarter. The government-backed IndiaAI Mission has handed Sarvam AI 4,096 NVIDIA H100 GPUs to develop a sovereign large language model, with three other Indian AI startups (Gnani.ai, SoketAI, and Gan.ai) also receiving state-backed compute support.
Why it matters: India is no longer just applying AI built elsewhere, it is now building the infrastructure layer. For students and professionals in India, this is a window: the domestic AI industry is scaling faster than the workforce filling it.
ClickUp cut 22% of its staff for AI agents. 2026 layoffs are nearly matching all of 2025.
ClickUp, the project management platform, cut roughly 22% of its workforce this week, explicitly citing AI agents that can now do the work those roles required. It is not alone: tech layoffs in 2026 are already tracking close to the full-year total from 2025. Analysts note that companies are no longer cutting costs to survive, they are cutting headcount to restructure around AI workflows.
Why it matters: This is the shift everyone talked about for two years, now showing up in payroll decisions. The roles most at risk are not low-skill ones, they are mid-tier coordination, writing, and support roles. Learning which AI tools apply to your work is no longer optional.
Anthropic is now the world's most valuable private AI company at $965 billion
Anthropic closed a $65 billion Series H round last week, pushing its post-money valuation to $965 billion, making it the most valuable private AI company in the world, surpassing OpenAI's $852 billion mark. The round was led by Altimeter Capital, Sequoia, and Dragoneer. Anthropic builds the Claude family of AI models and recently became the top AI platform in business adoption for the first time.
Why it matters: Less than two years ago, Anthropic was the cautious, safety-focused underdog. Now it is the most valued AI lab on the planet. The companies taking safety seriously are not losing the race, they may be winning it.
Autonomous agents are no longer a promise, they are a product
Microsoft's "Autopilots" are the clearest sign yet that AI is moving from a tool you call on to a system that acts on your behalf around the clock. Scout monitors your inbox. Other agents manage long workflows. The same architecture is coming from Google (Gemini Spark) and OpenAI. The question is no longer whether autonomous agents will arrive, it is whether people will know what to do when they realise their AI assistant has already sent that email, rescheduled that meeting, and written that report.
Colorado scales back its landmark AI law + Cognition AI raises $1B+
Colorado replaced its original AI Act (set to take effect June 30) with lighter disclosure-based rules under SB 189, pushing enforcement to January 2027 and removing the engineering-heavy requirements that had alarmed businesses. Separately, Cognition AI (the company behind software engineering agent Devin) closed a $1 billion+ Series D, with its revenue growing 13x in twelve months to a $492M annual run rate.