AI News Today June 1 2026: Top 10 Stories

AI News Today: June 1, 2026

June opens with a bang. SoftBank just wrote the single largest AI infrastructure check Europe has ever seen. An AI agent ran a real cyberattack from start to finish with zero human involvement. Humanoid robots are being field-tested in Ukraine. And in 7 days, Apple will attempt its biggest AI reset since Siri launched in 2011.

I've been tracking AI news daily and today feels like one of those sessions where multiple storylines converge at once. None of today's top stories were covered in yesterday's May 31 roundup. Here are the 10 that matter.

1. SoftBank Pledges €75 Billion ($87B) to Build Europe's Largest AI Data Centers in France

Breaking today: SoftBank Group has announced plans to invest up to €75 billion (approximately $87 billion) to develop and operate 5 gigawatts of AI data center capacity across France. It is the largest single AI infrastructure investment in European history, and SoftBank's biggest infrastructure commitment outside the United States.

The first phase involves €45 billion to deliver 3.1 GW of capacity across three sites in the Hauts-de-France region by 2031: Dunkirk, Bosquel, and Bouchain. Schneider Electric is a strategic partner at the Dunkirk site. The announcement is being made at the Choose France summit, reflecting personal diplomacy between Masayoshi Son and President Emmanuel Macron.

To understand the scale: Amazon's entire 2025 global capex was roughly $130 billion. SoftBank's France-only commitment is two-thirds of that. Europe has long struggled to attract AI infrastructure at this scale, primarily due to high energy costs. This deal changes the narrative.

My read: SoftBank is deeply tied to the AI boom through its Arm Holdings stake and OpenAI investment (more than $30 billion in, with $45 billion in gains as of March 2026). Son is backing Europe as a long-term AI hub, and France is making a credible play to become the continent's compute capital.

2. Sysdig Documents First-Ever LLM Agent Cyberattack, Database Exfiltrated in Under an Hour

This is the story that should keep every developer and CISO up at night. Sysdig's Threat Research Team documented the first publicly confirmed cyberattack driven entirely by a large language model (LLM) agent, observed on May 10, 2026.

Here's what happened: The attacker exploited CVE-2026-39987, a critical pre-authentication remote code execution flaw in Marimo, an open-source Python notebook platform. No exploit kit was needed. A single WebSocket request produced a full interactive shell. From there, an LLM agent took over: it harvested AWS credentials from the compromised host, used them to retrieve an SSH private key from AWS Secrets Manager, opened eight parallel SSH sessions through a bastion server, and fully exfiltrated a PostgreSQL database. The entire chain ran end-to-end in under one hour. The SSH bastion phase alone, which included dumping the full database schema and contents, completed in under two minutes.

The Sysdig team identified four markers that confirmed AI agent involvement: the attacker improvised a database dump with zero prior knowledge of the schema; a Chinese-language planning comment translated as 'see what else we can do' leaked directly into the command stream; every command used machine-readable formatting with delimiters; and 12 cloud API calls fanned across 11 distinct IPs in 22 seconds using Cloudflare Workers as a per-request egress pool to defeat source-IP detection.

The implication is stark. Static detection rules built around specific command patterns are now structurally obsolete. An LLM agent rewrites its approach for every target. Detection must shift to what the attacker is accomplishing (credential access, lateral movement, database exfiltration) not how. Patch Marimo immediately to version 0.23.0 or later.

3. Humanoid Robots Deployed to Ukraine Battlefield for First Time

CNBC reported this week that Foundation Future Industries, a San Francisco startup founded in 2024, sent two of its Phantom MK-1 humanoid robots to Ukraine for battlefield testing in February 2026. The company calls it the first known deployment of humanoid robots in a combat theater.

The Phantom MK-1 units performed logistics tasks in dangerous areas near the line of contact, including cargo delivery and evacuation runs. Foundation has received $24 million in Pentagon contracts from the US Army, Navy, and Air Force, and plans to scale production to thousands of units this year. The company also plans to send the upgraded Phantom MK-2 to Ukraine in 2026, with 'superhuman capabilities' and double the payload capacity.

The company has political ties: Eric Trump joined as chief strategic adviser after being an investor. Ukraine has conducted 7,495 robotics operations in a single month, with humanoids being one part of a broader AI-driven battlefield transformation. NATO-backed ARX Robotics is separately scaling autonomous ground vehicles to 1,800 units per year from a UK plant.

The honest take: 'tested in a combat zone' and 'deployed in combat' are different things. The Phantom MK-1 did logistics, not fighting. But the direction of travel is clear. Ukraine is the world's largest real-world test environment for autonomous military systems, and every tech startup with defense ambitions knows it.

4. WWDC 2026 in 7 Days: Apple's Gemini-Powered Siri Overhaul Is Coming

Apple's Worldwide Developers Conference kicks off Monday, June 8, at 10 a.m. Pacific. It is shaping up to be the most consequential AI event Apple has hosted since it first demonstrated Siri in 2011. Developer betas of iOS 27, iPadOS 27, macOS 27, and all other platforms will be available immediately after the keynote.

The centrepiece is a complete rebuild of Siri. According to Bloomberg's Mark Gurman, the new Siri runs on a custom model built on Google's Gemini technology, processed through Apple's Private Cloud Compute infrastructure rather than Google's servers directly. Apple and Google confirmed their multi-year AI collaboration in January 2026. The redesigned Siri integrates with the Dynamic Island, features a new 'Search or Ask' prompt, supports longer multi-turn conversations, and enables on-device processing for context-rich follow-ups even offline.

Apple registered the genai.apple.com subdomain on May 23, 2026, two weeks before the keynote. The company has never used 'GenAI' as a public-facing label before. It has also settled a $250 million class-action lawsuit over delayed Siri features promised at WWDC 2024.

I've been waiting for this for two years. If Apple actually delivers a Gemini-backed Siri that works on-device through Private Cloud Compute, the 1+ billion iPhone users who never touched ChatGPT or Claude will suddenly have frontier AI in their hands. That changes the consumer AI adoption curve more than any model release.

5. GitHub Copilot Token Billing Goes Live Today, Developer Backlash Builds

Starting today, June 1, 2026, all GitHub Copilot plans have transitioned to usage-based billing. Premium requests no longer exist as a fixed allowance. Instead, every plan includes a monthly pool of GitHub AI Credits based on its subscription price, consumed by actual token usage (input, output, and cached tokens) at the published API rate of whichever model you're using.

Plan prices are unchanged: Copilot Pro at $10/month, Pro+ at $39/month, Business at $19/user/month, and Enterprise at $39/user/month. But what's changed is predictability. A long agentic coding session now burns dramatically more credits than a quick chat question. GitHub CPO Mario Rodriguez's framing, that 'a short chat question can cost the user just as much as an autonomous coding session,' explains the shift, but doesn't make the pill easier to swallow.

TechCrunch described developer reactions as 'consternation,' with some calling it 'what a joke.' The concern is legitimate for smaller teams and indie developers who relied on Copilot's flat-rate pricing model to budget predictably. The new model is fair by cost-alignment logic but brutal for anyone who was heavily using agentic Copilot features under the old regime.

The bigger story: this marks the moment AI coding tools leave the promotional phase and enter the managed-infrastructure phase. The unlimited-feeling AI coding era is over. Welcome to metered AI compute.

Another pricing shift hits today: OpenAI's 2x launch promotion for its $100/month ChatGPT Pro tier expired on May 31, 2026. This means Codex usage on the Pro 5x plan (the $100 tier) drops from the promotional 10x Plus level back down to the standard 5x Plus from today onward.

For developers on the $100 Pro tier who have been relying on the double capacity for high-volume Codex sessions, this is a halving of effective usage without any price change. The $200 Pro tier (Pro 20x) was already on a separate promotional track and maintained its usage levels. OpenAI Codex had more than 3 million weekly active users as of April 2026, making this a change that affects a meaningful developer population.

Paired with the GitHub Copilot changes also effective today, June 1 marks the formal end of the 'unlimited AI coding' era. Every major AI coding tool is now either metered by tokens or metered by compute credits. Developers who want predictable pricing need to budget accordingly.

7. Novo Nordisk and OpenAI Strike Enterprise AI Deal for Drug Discovery

Danish pharmaceutical giant Novo Nordisk announced a strategic partnership with OpenAI on April 14, 2026 to integrate AI end-to-end across its business, from drug discovery and clinical trials to manufacturing, supply chain, and commercial operations. Full deployment is targeted by end of 2026.

The partnership will use advanced AI to analyze complex biological datasets, identify promising drug candidates, and reduce R&D timelines. CEO Mike Doustdar framed the goal as enabling Novo to 'analyse datasets at a scale that was previously impossible.' OpenAI CEO Sam Altman said AI can help 'people live better, longer lives' in life sciences.

Context matters here: Novo Nordisk is in an intense race with Eli Lilly for dominance in the GLP-1 weight-loss drug market, where Lilly has overtaken it. Lilly has signed 16 AI deals since 2025 totaling billions, including a $2.75 billion Insilico Medicine partnership. The AI drug discovery market is projected to grow from roughly $3-8 billion in 2026 to over $25 billion by 2035.

This is one of the clearest examples of AI moving from a tech product into the physical world of human health. If AI can genuinely compress multi-year drug development cycles, the downstream effects are enormous.

8. DeepSeek Makes 75% V4-Pro Price Cut Permanent, Escalating the Inference War

DeepSeek has made its 75% price cut on V4-Pro permanent, a move that turns what looked like a promotional discount into a structural price signal for the entire AI inference market. The reduction escalates what AI Weekly has called the inference war, where Chinese labs use aggressive pricing to challenge US frontier model dominance.

DeepSeek V4, released in early May 2026, demonstrated competitive performance on non-Nvidia hardware (including Huawei's Ascend chips), and V4-Pro at its new permanent pricing undercuts most US frontier models on a cost-per-token basis by a significant margin. This pricing pressure has been one of the catalysts behind ByteDance's capex push to build domestic AI infrastructure, as detailed in yesterday's roundup.

The inference war has a simple dynamic: as US labs race to the frontier on capability, Chinese labs race to the bottom on price. Developers building on AI APIs are caught in the middle, choosing between cutting-edge capability and cost efficiency. DeepSeek's permanent price cut makes that trade-off sharper than ever.

9. Apple Registers genai.apple.com Ahead of WWDC — Biggest Brand Signal Yet

On May 23, 2026, Apple quietly registered the subdomain genai.apple.com, two weeks before its WWDC 2026 keynote. The page does not yet resolve, but the timing and labeling are significant: Apple has never used the phrase 'GenAI' as a public-facing label before, consistently preferring 'Apple Intelligence' since 2024.

MacRumors leaker Aaron Perris spotted the registration, and multiple outlets have since confirmed it. The prevailing interpretation is that Apple is building a public-facing marketing hub for its Gemini-powered Siri rollout, possibly branded differently from its broader Apple Intelligence suite. The Dynamic Island integration for the new Siri, confirmed by MacRumors, would make Siri feel more like a persistent AI layer than a voice button.

Apple also settled a $250 million class-action lawsuit on May 5, 2026 over delayed Siri features promised at WWDC 2024. The settlement serves as a meaningful commitment: Apple now has legal and financial accountability, not just brand pressure, for delivering on AI promises.

10. Anthropic and Gates Foundation Pledge $200M to Put Claude in Global Health

Anthropic and the Bill and Melinda Gates Foundation have pledged $200 million to integrate Claude into global health programs, according to AI Weekly's May 26 coverage. The initiative aims to deploy Claude-powered tools inside government health systems, NGO operations, and clinical workflows in low-and-middle-income countries.

The partnership is a direct extension of Anthropic's social impact strategy alongside its commercial growth. Anthropic's $47 billion annualized revenue run rate is built almost entirely on enterprise and developer usage, but this pledge signals a parallel track: AI for public good at scale.

The Gates Foundation has historically been one of the most effective philanthropic operators in global health. Combining that distribution capability with Claude's language and reasoning abilities creates a potentially significant tool for healthcare workers who lack access to specialists or diagnostic resources. No product launch dates have been announced.

Frequently Asked Questions

Q: What is the SoftBank France AI investment and how big is it?

SoftBank Group announced on May 30-31, 2026 that it plans to invest up to €75 billion (approximately $87 billion) to build 5 gigawatts of AI data center capacity in France. The first phase involves €45 billion for 3.1 GW of capacity in Hauts-de-France by 2031, across sites in Dunkirk, Bosquel, and Bouchain. It is the largest AI infrastructure investment in European history.

Q: What was the Sysdig LLM agent cyberattack?

On May 10, 2026, Sysdig's Threat Research Team documented the first publicly confirmed cyberattack driven entirely by an LLM agent. The attacker exploited CVE-2026-39987, a critical remote code execution flaw in Marimo (an open-source Python notebook), then an AI agent autonomously harvested credentials, retrieved SSH keys from AWS Secrets Manager, and fully exfiltrated a PostgreSQL database. The entire attack chain completed in under one hour.

Q: What will Apple announce at WWDC 2026?

WWDC 2026 begins June 8, 2026 at 10 a.m. Pacific. Apple is expected to unveil iOS 27, iPadOS 27, macOS 27, watchOS 27, tvOS 27, and visionOS 27, all focused on AI. The headline feature is a rebuilt Siri powered by a custom model based on Google's Gemini technology, processed through Apple's Private Cloud Compute infrastructure. Siri will integrate with the Dynamic Island and support longer, context-rich conversations.

Q: What changed with GitHub Copilot billing on June 1?

On June 1, 2026, GitHub Copilot switched from premium request-based usage to token-based billing using GitHub AI Credits. Plan prices stayed the same (Pro at $10/month, Business at $19/user/month), but costs now depend on actual token consumption rather than a fixed request allowance. Heavy agentic coding sessions consume significantly more credits than simple chat interactions.

Q: What happened to OpenAI Codex Pro pricing today?

The 2x launch promotion for OpenAI's $100/month ChatGPT Pro tier expired on May 31, 2026. From June 1, the Pro 5x tier reverts to standard 5x Plus usage instead of the promotional 10x Plus. Users on the $100 Pro tier who relied on doubled Codex capacity will see their effective usage halve without a price reduction.

Q: What is Foundation Future Industries and the Phantom robot?

Foundation Future Industries is a San Francisco startup founded in 2024 that builds dual-use humanoid robots for industrial and military applications. Its Phantom MK-1 robot was deployed to Ukraine in February 2026 for logistics testing near the front line, marking what the company describes as the first known deployment of humanoid robots in a combat theater. The company has received $24 million in Pentagon contracts.

Q: What is DeepSeek V4-Pro and why does the price cut matter?

DeepSeek V4-Pro is a frontier AI model from Chinese lab DeepSeek, released in May 2026. DeepSeek has made its initial 75% price cut permanent, positioning V4-Pro as one of the lowest-cost frontier models available. The permanent cut signals that Chinese AI labs are using price as a structural competitive weapon against US frontier model providers, not just a promotional tool.

Q: What is the Anthropic and Gates Foundation $200M AI health initiative?

Anthropic and the Bill and Melinda Gates Foundation pledged $200 million to integrate Claude into global health programs, with a focus on deploying AI tools inside government health systems, NGO operations, and clinical workflows in low-and-middle-income countries. The initiative was reported by AI Weekly in late May 2026. No specific product launch dates have been announced.

AI is moving faster than a news cycle. The best way to keep up is consistent daily learning, not occasional catch-up sessions.

Learn AI in 5 minutes a day on Unrot — the microlearning app for people who want to stay ahead without burning out.