Dixon AI Newsletter May 2026

Aug 10

11 min read

Welcome to the Dixon AI Newsletter which summarises selected AI stories and releases you might have missed. This issue covers AI news from 25th April 2026 to 28th May 2026.

The AI Macro Environment and Enterprise Insights

Anthropic strikes Colossus compute deal with SpaceX, raising Claude usage limits: Anthropic signed a deal to take all compute capacity at SpaceX's Colossus 1 data centre in Memphis, gaining access to over 220,000 NVIDIA GPUs and more than 300 megawatts of capacity within a month. The expanded compute let Anthropic double Claude Code's five-hour rate limits for Pro, Max, Team and Enterprise plans, remove peak-hours reductions for Pro and Max users, and raise API rate limits for Opus models. Anthropic also holds compute deals with Amazon, Google, Microsoft and NVIDIA. See (Anthropic).
Anthropic's revenue and valuation trajectory reframes frontier lab economics: Anthropic closed a $65 billion Series H on 28th May at a post-money valuation of $965 billion, likely its last private round before an IPO reportedly targeted for late 2026. The company's first operating profit (~$559m) is projected for Q2, with run-rate revenue reportedly crossing $47 billion, up from $9 billion at end of 2025 and overtaking OpenAI. Enterprise customers spending over $1m a year doubled from 500 to 1,000 in under two months, and data shows 1 in 5 of its businesses now pay for Anthropic, up from 1 in 25 a year ago. See (TechCrunch).
OpenAI Codex hits 4m weekly developers as Dell partnership opens on-premises route: OpenAI confirmed Codex passed 4 million weekly developers by 21st April, up from roughly 600,000 at the start of the year (a near 7x increase). Codex enterprise clients now include Cisco, Datadog, Dell and NVIDIA, with usage extending beyond coding into reporting, lead qualification and cross-system coordination. See (OpenAI).
Pope Leo XIV publishes the first Catholic encyclical on AI: Magnifica humanitas: On safeguarding the human person in the time of artificial Intelligence was unveiled at the Vatican on 25th May. Anthropic co-founder Chris Olah spoke at the presentation, acknowledging that frontier labs face commercial and geopolitical pressures that can conflict with doing the right thing, and called for outside moral voices. The document is the most significant religious-institutional intervention on AI ethics to date and is likely to shape how faith-aligned organisations, charities and educational bodies approach AI policy. See (Anthropic).
US pre-release AI model reviews expanded: Google, Microsoft and xAI agreed to let the Commerce Department's Center for AI Standards and Innovation (CAISI) conduct pre-deployment evaluations of new frontier models, joining OpenAI and Anthropic. CAISI has completed 40 model reviews since 2024. The arrangement signals that voluntary government safety review is hardening into expected practice for any lab planning a major release. See (The Verge).
Meta sued by major publishers over Llama training data: Macmillan, McGraw Hill, Elsevier, Hachette and Cengage, alongside author Scott Turow, filed a class action against Meta alleging Llama was trained on books pirated from LibGen and Sci-Hub. The suit calls it "one of the most massive infringements of copyrighted materials in history" and demands a full list of training works. The case sharpens the precedent risk for any enterprise relying on frontier models in regulated content workflows. See (The Verge).
Google DeepMind workers move to unionise over military contracts: Staff at DeepMind's London HQ voted 98% in favour of recognition through the CWU and Unite, with at least 1,000 employees behind the bid. Workers want Google to commit not to develop weapons technologies and to allow ethical opt-outs. The vote is the most significant labour action yet inside a frontier lab and creates a template other AI workforces may follow. See (The Verge).
OpenAI and Anthropic launch parallel enterprise deployment vehicles: Within minutes of each other, OpenAI announced a $10bn JV with TPG, Brookfield, Advent and Bain to accelerate enterprise rollout, and Anthropic launched its own services firm with Blackstone, Hellman & Friedman and Goldman Sachs. Both target mid-market customers (community banks, manufacturers, regional health systems) that lack internal AI capability. Implementation, not models, is now the competitive front. See (Anthropic).
NYT Tech Guild files unfair labour charge over AI monitoring: The 700-member union accused the New York Times of using DX (engineering productivity tracking) and Glean (internal search) in disciplinary proceedings without bargaining. The 1,500-person editorial Guild filed parallel charges over the paper's refusal to share AI plans. The case will shape how organisations introduce AI productivity tooling into unionised workforces. See (The Verge).
Universal Music music-rights deals reshape the AI music economy: UMG and TikTok renewed their licensing agreement with explicit terms on removing unauthorised AI-generated music and improving artist attribution. UMG also signed with Spotify allowing licensed fan-made AI covers and remixes with revenue share for participating artists. The deals offer a template for how rights holders and platforms will handle generative AI as enforcement tightens. See (TechCrunch).
Coinbase cuts 14% of staff in "AI-native" restructuring: Brian Armstrong told staff the reduction was part of a strategic shift to become an AI-native company, with AI reshaping how the exchange operates and is staffed. The framing matters: it is one of the first large public-company layoff announcements explicitly attributed to AI restructuring rather than market conditions, and likely a leading indicator. See (Brian Armstrong on X).
Anthropic and Gates Foundation commit $200m to global development AI: The four-year programme covers grants, Claude credits and engineering support across global health, life sciences, education and economic mobility. It is run through Anthropic's Beneficial Deployments team, which also produces public-good evaluation benchmarks and datasets. This is the most concrete sign yet that frontier labs see development institutions as strategic distribution partners. See (Anthropic).
Brockovich launches crowdsourced AI data centre map: Environmental campaigner Erin Brockovich has launched a public reporting site mapping operational and proposed AI data centres across the US, with 2,716 reports logged at launch. Texas dominates, with 297 reports from Sulfur Springs alone where MSB Global plans a gigawatt facility on 1,600 acres. Community opposition over water, electricity and pollution is now an organised constraint on siting decisions. See (Engadget).

New and updated AI model and application releases

OpenAI

GPT-5.5 Instant: new ChatGPT default model. 52.5% fewer hallucinated claims than GPT-5.3 Instant on high-stakes prompts, plus better image analysis and personalisation. Introducing GPT-5.5 Instant
Trusted Contact: optional ChatGPT safety feature letting adults nominate someone to be alerted if self-harm concerns are detected. Developed with the American Psychological Association. Introducing Trusted Contact in ChatGPT
Realtime Voice Models: new API voice models for reasoning, translation and transcription, enabling more natural voice experiences. Advancing voice intelligence

Anthropic

Claude Opus 4.8: latest flagship model released 28th May, with stronger performance on coding, agentic tasks and professional work, and improved consistency for long-running tasks. Introducing Claude Opus 4.8
Claude for Excel, PowerPoint and Word generally available; Outlook in public beta. Single conversation thread carries context across all four apps. Collaborate with Claude across Excel, PowerPoint, Word and Outlook
Self-hosted sandboxes and MCP tunnels for Claude Managed Agents: Enterprises can run agent tool execution on their own infrastructure or via Cloudflare, Daytona, Modal or Vercel. New in Claude Managed Agents
Claude Compliance API and 28 new security and compliance integrations: including Cloudflare, CrowdStrike, Microsoft Purview, Okta, Palo Alto Networks, Datadog and Zscaler. Claude works with more security and compliance tools
Project Glasswing: Claude Mythos Preview found over 10,000 high or critical-severity software vulnerabilities in a month working with around 50 partners. Project Glasswing initial update
Higher usage limits via SpaceX Colossus deal: rate limits doubled for Pro, Max, Team and Enterprise after Anthropic took all compute capacity at the Colossus 1 data centre. Higher usage limits for Claude and a compute deal with SpaceX

Google

Gemini 3.5 Flash: first in the new 3.5 series, generally available via the Gemini App, Antigravity, Gemini API, AI Studio and Android Studio. Outperforms 3.1 Pro on coding and agentic benchmarks at roughly 4x the speed. Gemini 3.5 Flash
Gemini Omni: new multimodal tool accessible within Gemini combining reasoning with creation, starting with video. Google I/O 2026 keynote
Gemini 3.1 Flash-Lite: generally available on the Gemini Enterprise Agent Platform. Designed for ultra-low latency, high-volume agentic tasks. Gemini 3.1 Flash-Lite is now generally available
Antigravity 2.0: upgraded agent-first development platform with a new CLI, subagent orchestration and built-in sandboxing. Antigravity 2.0
100+ I/O announcements: including Pomelli AI brand agents, Stitch design updates, Android Halo status-bar agent, voice updates to Gmail/Docs/Keep, and expanded SynthID/C2PA watermarking. 100 things we announced at Google I/O 2026

Microsoft

GPT-5.5 and Opus 4.8 in Microsoft 365 Copilot: both Instant and Thinking GPT-5.5 variants now available via the model picker along with Opus 4.6. Available today: GPT-5.5 Instant in Microsoft 365 Copilot
Microsoft 365 Copilot redesign: rolled out 28th May with a new task-aware workspace, unified Copilot entry point across Word, Excel, PowerPoint, Outlook and Teams, pinned conversations, progressive-disclosure controls, and reported load times more than twice as fast. Introducing a new design for Microsoft 365 Copilot
Work IQ rollout: the intelligence layer beneath the redesign, grounding Copilot in cross-app context (files, emails, meetings, chats, relationships and organisational signals) and powering the model picker that lets users choose between OpenAI and Anthropic models in Excel and Word. Work IQ now also exposes a REST API and CLI via Copilot Studio for embedding into existing workflows, with agent-to-agent communication generally available. What's new in Microsoft 365 Copilot, May 2026 and What's new in Copilot Studio, May 2026
MAI-Image-2.5: in-house text-to-image model, debuted at number three on the Arena leaderboard. Available on Arena now, MAI Playground and Azure Foundry within two weeks. MAI-Image-2.5 launches at No. 3 on Arena

Notable Other Model and Application Releases

xAI Grok Build: beta tool for SuperGrok and X Premium+ subscribers with Plan Mode, Imagine for image and video generation, and a CLI for building agents and orchestrators. xAI Grok Build announcement
ElevenLabs Music v2: updated AI music generation model with improved vocals, instrumentation and multilingual support. ElevenLabs Music v2
Adobe productivity agent for Acrobat: chat with PDFs and generate presentations, podcasts, blogs and social posts. Available in Acrobat Express and Acrobat Studio. Adobe's new productivity agent
Perplexity Personal Computer for Mac: macOS app letting Perplexity's agents run continuously and autonomously across local files, native Mac apps and over 400 connectors. Personal Computer is available to all Mac users
Runway Aleph 2.0: edit a single video frame and propagate the change consistently across the entire clip. Runway Aleph 2.0
Runway MCP and Ideogram MCP: both launched MCP integrations bringing image and video generation directly into Claude, ChatGPT, Cursor and Replit. Runway MCP

From Dixon AI

AI Drop In: one-hour, live sessions held twice a week.
- Join here: https://www.dixonai.com/ai-drop-in
  - w/c 1st June: What Happened in AI in May
  - w/c 8th June: Getting started with Claude Cowork
  - w/c 15th June: Bots and Get Cody
  - w/c 22nd June: Creating presentations with AI
  - w/c 29th June: What Happened in AI in June
- Please note that June will be the last month that AI Drop-in sessions will be freely available to access via the our website and from July onwards access will be limited to clients via the Dixon AI Hub
AI Courses:
- We are now running online three-week AI Literacy Essentials courses and four-week Masterclasses on Claude, ChatGPT and Copilot365 every month
- For more information about our online courses, or to sign up to a course, please visit https://www.dixonai.com/training

A Word from Rob - May 2026

Speaking to a room full of corporate leaders in New York this month, Ethan Mollick said something that's been quoted everywhere since. Nobody, he argued, actually has the playbook for this. Not the AI labs, not the consultancies, not the CEOs. We're all making it up as we go along, and anyone who tells you they've cracked it is selling you something (Mollick, reported in Fortune, 2026).

The point underneath the soundbite is the important one, and he's right about it. AI making individuals more productive doesn't automatically make an organisation more productive. Closing that gap takes organisational change, not just better tools, and it's precisely where most organisations are stuck.

The labs are making the same point with their money. This month OpenAI and Anthropic each launched enterprise deployment ventures within days of one another, both backed by major private equity, both aimed at the mid-market organisations that can't yet embed AI themselves (Axios, 2026). Coinbase, meanwhile, cut 14 per cent of its workforce, with its CEO naming AI directly and describing how a small, focused team can now move at a pace that's reshaping how the business is run (Fortune, 2026). Implementation, not models, is now the game.

So here's where I'd gently push back. When we talk about a "playbook for AI", what we're really talking about is a playbook for accelerating innovation inside an organisation. And innovation has a settled meaning: taking something new, getting genuinely good at it, building the capability to use it well, then applying it to business as usual so it earns its keep. Put that way, the picture changes, because there are plenty of playbooks for doing exactly that. They live in strategic change, in innovation theory, in how good entrepreneurs actually work, and they've been studied and refined for decades.

The most useful shift a leader can make right now is to stop treating this as an AI problem and start treating it as an innovation problem. The lessons aren't going to come from the technology. They come from change, innovation, and entrepreneurship, which, as it happens, was my world for years before AI came along.

That's what we set out to do in The AI Transformation Playbook. Its opening section leans on the people who mapped this ground long before us: Lewin on how change actually takes hold, Kotter on leading it, Whittington on strategy as something people do rather than own, and Sarasvathy on how entrepreneurs build under uncertainty through effectuation. None of it is new. It's established discipline, applied honestly to a genuinely new situation. You can get a copy here: https://www.amazon.co.uk/dp/1919400508.

I'm not claiming a perfect playbook. AI is too new for anyone to have that, and Ethan's right to be wary of those who say otherwise. But across more than fifty organisations over the past two years, the Dixon AI team and I have refined an approach that is a very long way from nothing. That refinement, far more than the book, is the real work.

If you'd like to talk about what that looks like for your organisation, let's talk.

Rob

References

Axios, 2026. OpenAI and Anthropic partner with private equity. [online] Axios, 4 May. Available at: https://www.axios.com/2026/05/04/openai-anthropic-private-equity-enterprise-business [Accessed 31 May 2026].

Dixon, R.J., 2025. The AI Transformation Playbook: A Step-by-Step Guide to Becoming an AI-Enabled Organisation. Manchester: Dixon AI Press. Available at: https://www.amazon.co.uk/dp/1919400508

Fortune, 2026. Coinbase didn't just lay off 14% of its staff due to AI. It replaced managers with 'player-coaches' and turned its org chart upside down. [online] Fortune, 5 May. Available at: https://fortune.com/2026/05/05/coinbase-layoffs-14-of-employees-ai-tech-ai-job-anxiety-crypto/ [Accessed 31 May 2026].

Mollick, E., 2026. Remarks to corporate leaders, New York, May 2026. Reported in: 'Nobody knows anything' and 'this time is different': the phrases that define, and haunt, the AI economy. [online] Fortune, 25 May. Available at: https://fortune.com/2026/05/25/nobody-knows-anything-this-time-is-different-ethan-mollick-no-playbook-ai/ [Accessed 31 May 2026].

Dixon AI Newsletter May 2026

The AI Macro Environment and Enterprise Insights

Recommended Reading, Listening and Book Releases

From Dixon AI

A Word from Rob - May 2026

References

Related Posts

Comments