AI News Today May 31 2026: Top 10 Stories

AI News Today: May 31, 2026

I've been tracking AI news daily for two years, and I can say with confidence: this past 72 hours is one of the most event-dense periods I've seen. Anthropic nearly hit a $1 trillion valuation. GitHub Copilot pricing is changing tomorrow. ByteDance is building an AI infrastructure war chest the size of a small country's economy. And the man who created AlphaGo just raised $1.1 billion for a startup with zero products.

Here are the 10 biggest AI news stories from the last 24 hours, ranked by importance. No filler. No hype. Just what actually matters.

1. Anthropic Raises $65B, Launches Claude Opus 4.8, Hits $965B Valuation

On May 28, 2026, Anthropic did two massive things in one day: raised $65 billion in Series H funding at a $965 billion post-money valuation, and launched Claude Opus 4.8.

The funding round was co-led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN as co-leads. Strategic hardware partners Samsung, SK Hynix, and Micron also participated. Amazon's previously committed $5 billion is included in the $15 billion hyperscaler portion.

At $965 billion, Anthropic is now worth more than OpenAI ($852B after its March 2026 round). Its annualized revenue run rate crossed $47 billion earlier in May, up from $14 billion in February. That's one of the fastest revenue ramps in tech history.

Claude Opus 4.8 launched on the same day, just 41 days after Opus 4.7. Key improvements: it's approximately 4x less likely to let faulty code pass without flagging the issue, scored 74.2% on Terminal-Bench 2.1 (up 8.4% from 4.7), and is stronger on agentic tasks, financial analysis, and computer use. Pricing stays the same: $5 per million input tokens, $25 per million output tokens. Fast Mode is now 3x cheaper than it was for 4.7.

My take: Anthropic's revenue tripling in three months on the back of Claude Code is remarkable. This is no longer a safety-first challenger story. This is a full enterprise infrastructure company that also happens to care about alignment.

2. GitHub Copilot Switches to Token-Based Billing Tomorrow (June 1)

If you use GitHub Copilot, your pricing structure is changing tomorrow. GitHub announced all Copilot plans will move to usage-based billing on June 1, 2026, replacing the premium-request system with token-based AI Credits.

Plans and base prices stay the same: Copilot Pro at $10/month, Pro+ at $39/month, Business at $19/user/month, Enterprise at $39/user/month. But instead of a fixed request allowance, each plan now comes with a monthly AI Credit pool equal to the plan price, consumed based on actual token usage (input + output + cached).

Why this matters: heavy Copilot users doing multi-hour agentic coding sessions could face significantly higher bills. TechCrunch coverage described developer reactions as consternation, with some calling it 'what a joke.' GitHub CPO Mario Rodriguez defended the shift by pointing out that 'a short chat question can cost the user just as much as an autonomous coding session lasting several hours.'

I see both sides. The flat-rate model was always subsidized. But losing price predictability is real friction for small teams and indie developers. This is the moment AI coding tools stop feeling like utilities and start behaving like cloud services.

3. ByteDance Mulls $70B AI Capex to Challenge US Hyperscalers

Bloomberg reported this week that ByteDance is discussing capital expenditures of up to $70 billion in 2026 as it builds out AI data centers and infrastructure. The company plans to fund most of it from the roughly $50 billion profit it earned in 2025.

To put that in perspective: Amazon's 2025 capex was about $130 billion, and Alphabet's 2026 capex is projected at $185 billion. ByteDance at $70B would make it one of the most aggressive AI infrastructure spenders anywhere, and more aggressive than Tencent ($12B) or Alibaba ($19B) by a wide margin.

ByteDance has also reportedly struck a deal with Qualcomm to develop custom AI chips for its data centers. Separately, Huawei's Ascend 950 chips are seeing surging demand from Chinese companies including ByteDance, partly because US export controls limit access to Nvidia's most advanced processors.

The story here isn't just money. It's infrastructure as a geopolitical weapon. China's AI infrastructure race is constrained by chip access but ByteDance is betting that sheer scale of spending and domestic supply chain investment can close the gap

4. OpenAI Files Confidential IPO Paperwork, Eyes September Listing

OpenAI is preparing to file a confidential draft registration statement with the SEC, with Goldman Sachs and Morgan Stanley leading the process. The target: a public debut in September 2026, at a valuation above $1 trillion.

The company's last disclosed private valuation was $852 billion after its $122 billion funding round in March 2026. Monthly revenue is currently $2.6 billion. CFO Sarah Friar has previously confirmed that the company will reserve IPO shares for retail investors.

Anthropic is separately targeting an October 2026 IPO at a valuation above $900 billion (now potentially higher given the $965B Series H). SpaceX already filed its public S-1 on May 20, 2026. If both OpenAI and Anthropic list within months of each other, the AI sector will have its first major public market transparency test.

This matters for everyone in AI. Once these companies are public, we'll finally have real numbers on AI revenue, unit economics, and cost structures. Right now we're all working from press releases and analyst estimates.

5. DeepMind CEO: AI Is a 'Species-Level Transition'

At a talk at Stanford's Graduate School of Business on May 29, 2026, Google DeepMind CEO Demis Hassabis described AI as entering a period unlike any previous technological shift: a 'species-level transition' that leaves humanity with 'little margin for error' over the next decade.

Hassabis said AI is currently in the 'foothills of the singularity,' with the technology advancing approximately 10 times faster than the Industrial Revolution. He called for increasing international coordination on AI regulation within the next 5-10 years.

The DeepMind CEO has consistently occupied a middle position: deeply optimistic about AI's potential to solve science's biggest problems, while genuinely concerned about the risks of moving too fast without governance infrastructure. I find this a more credible position than either pure accelerationism or pure doom.

6. EU AI Act Presses Anthropic Over Claude Mythos Access

The EU is stepping up pressure on Anthropic to provide access to Claude Mythos Preview, its powerful cybersecurity-focused model. The European Commission has held multiple technical meetings with Anthropic but has not yet gained model access.

EU AI Act enforcement provisions for general-purpose AI models take effect on August 2, 2026. After that date, the EU AI Office will have formal authority to require access. An EU spokesperson said: 'Once the enforcement powers of the AI Office start in August 2026, we will ensure to receive, if needed, model access.'

OpenAI has separately offered the European Commission access to GPT-5.5-Cyber for regulatory review, positioning itself as the cooperative alternative. Separately, the EU has also delayed broader AI Act compliance rules, with provisions originally due August 2026 for standalone AI systems now pushed back to December 2027.

The Mythos access dispute is a preview of a much bigger fight coming: who gets to inspect the most powerful AI systems, on what terms, and with what authority. Europe is finding out that writing regulation is easier than enforcing it against American companies.

7. David Silver's Ineffable Intelligence Raises $1.1B Seed Round

Ineffable Intelligence, the London-based AI lab founded by former DeepMind reinforcement learning lead David Silver, has raised $1.1 billion at a $5.1 billion valuation in what is reportedly the largest seed round ever in Europe. The round was co-led by Sequoia and Lightspeed, with participation from Nvidia, Google, Index Ventures, DST Global, and the UK's Sovereign AI Fund.

Silver is the researcher behind AlphaGo and AlphaZero, programs that mastered board games through self-play without human examples. Ineffable Intelligence aims to scale that same reinforcement-learning approach to build a 'superlearner': an AI that discovers knowledge through experience rather than human-generated training data.

The company has no product, no revenue, and no public roadmap. First model benchmarks are expected by late 2026. What they do have: arguably the most credible RL researcher alive, and a very compelling contrarian thesis to the 'just scale LLMs' orthodoxy that dominates right now.

Silver has also pledged to donate 100% of his personal proceeds from Ineffable to high-impact charities via Founders Pledge.

8. Claude Mythos Public Release Coming 'Soon'

Fortune reported alongside Anthropic's Series H announcement that the company is planning to more widely release models on par with Claude Mythos Preview, its advanced cybersecurity-capable model currently limited to approximately 40 partner organizations via Project Glasswing.

Anthropic's own evaluations show Mythos Preview achieved its highest-ever scores on alignment measures, and Opus 4.8's misaligned behavior rates are 'substantially lower than Opus 4.7' and 'comparable to Claude Mythos Preview.' That language suggests the gap between Mythos and the public frontier is narrowing.

On Project Glasswing: in approximately 30 days, Mythos Preview and around 50 partner organizations identified more than 10,000 high or critical-severity vulnerabilities across open-source projects. Of those, 1,726 were confirmed true positives, with 1,094 classified as high or critical severity. Cloudflare alone found 2,000 bugs in its systems using Glasswing access.

The Mythos public release is one of the most anticipated events in AI right now. A model that can find zero-days autonomously, at scale, will fundamentally change cybersecurity if it reaches general availability.

9. Huawei Expects AI Chip Revenue to Hit $12B in 2026

Huawei expects its AI chip revenue to reach $12 billion in 2026, a 60% increase year-over-year, as companies including ByteDance and Alibaba race to secure supply of its Ascend 950PR chips. The surge in demand follows US export controls that restrict access to Nvidia's most advanced accelerators for Chinese buyers.

DeepSeek's V4 model release demonstrated competitive performance on non-Nvidia hardware, accelerating the shift toward Huawei's Ascend ecosystem among Chinese AI companies. ByteDance's capex plans include a significant portion earmarked for domestic AI chips as it reduces dependence on Nvidia.

The geopolitical dimension: US export controls were designed to slow China's AI progress. Instead, they've accelerated domestic chip development and created a multi-billion dollar revenue opportunity for Huawei. This is a classic case of a policy having the opposite of its intended long-term effect.

10. OpenAI Retiring o3 and GPT-4.5 from ChatGPT

OpenAI has announced model retirements: o3 will be retired from ChatGPT on August 26, 2026 (following a 90-day sunset period), and GPT-4.5 will be retired from ChatGPT on June 27, 2026 after a 30-day sunset. Both retirements affect ChatGPT only; API access continues unchanged.

Both models are currently available to paid ChatGPT users only via model settings. The retirements reflect OpenAI's ongoing model lifecycle management as newer models like GPT-5.5 and o-series successors take over the product layer.

For developers: check your ChatGPT integrations if you're relying on specific model versions. The API versions remain available, so production workloads are unaffected.

Frequently Asked Questions

Q: What is Claude Opus 4.8 and how is it different from Opus 4.7?

Claude Opus 4.8, released May 28, 2026, is Anthropic's latest flagship model. It is approximately 4x less likely than Opus 4.7 to let faulty code pass without flagging the issue. It scored 74.2% on Terminal-Bench 2.1 (an 8.4% improvement) and 4.9% higher on SWE-Bench Pro. Pricing remains the same at $5 per million input tokens and $25 per million output tokens.

Q: How much did Anthropic raise and what is its current valuation?

Anthropic raised $65 billion in its Series H round on May 28, 2026, at a $965 billion post-money valuation. The round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. Its annualized revenue run rate crossed $47 billion earlier in May 2026, up from $14 billion in February.

Q: When does GitHub Copilot switch to usage-based billing?

GitHub Copilot transitions to usage-based billing on June 1, 2026. All plans switch from premium-request counting to a token-based AI Credits model. Base subscription prices stay the same, but costs are now tied to actual token consumption rather than a fixed request allowance.

Q: What is Ineffable Intelligence?

Ineffable Intelligence is a London-based AI startup founded in late 2025 by David Silver, former reinforcement learning lead at Google DeepMind. In April 2026, it raised $1.1 billion at a $5.1 billion valuation in the largest seed round ever in Europe. The company aims to build a 'superlearner' that acquires knowledge through reinforcement learning without human-generated training data.

Q: When is the Claude Mythos public release?

As of May 31, 2026, Claude Mythos Preview remains restricted to approximately 40 partner organizations via Anthropic's Project Glasswing cybersecurity initiative. Anthropic has signaled a broader release is coming 'soon,' with Fortune reporting this alongside the Series H announcement on May 28, 2026. No exact date has been confirmed.

Q: What is OpenAI's IPO timeline in 2026?

OpenAI is preparing to file a confidential draft IPO prospectus with the SEC, with Goldman Sachs and Morgan Stanley advising. The company is targeting a public debut in September 2026 at a valuation above $1 trillion. Its last private valuation was $852 billion after its $122 billion funding round in March 2026. No confirmed filing date or ticker has been announced.

Q: How much is ByteDance spending on AI in 2026?

Bloomberg reported on May 27, 2026 that ByteDance is discussing capital expenditures of up to $70 billion in 2026 for AI data centers and infrastructure. The company plans to fund much of this from approximately $50 billion in profit earned in 2025. ByteDance has also separately struck a deal with Qualcomm to develop custom AI chips.

Q: What did Demis Hassabis say about AI at Stanford?

At Stanford GSB on May 29, 2026, Google DeepMind CEO Demis Hassabis called AI a 'species-level transition' unlike previous technological shifts, and said humanity has 'little margin for error' over the next decade. He said AI is advancing approximately 10 times faster than the Industrial Revolution and called for international coordination on AI regulation within 5-10 years.

AI moves fast. Understanding it daily beats trying to catch up on weekends.

Learn AI in 5 minutes a day on Unrot — the microlearning app built for people who don't have hours to spare.