This week’s AI news arrived in every direction at once: court drama, government mandates, model launches, and a compute deal that sounds like science fiction. The through-line, as always, is speed. The decisions being made this week will define the competitive landscape for years to come, and the businesses that understand them now hold a material advantage over those still watching from the sidelines.
THIS WEEK’S TOP 3 MUST-KNOW CONCEPTS
Half the Government, Run by AI
GOV & AGENTIC AI
The UAE has announced a plan to place 50% of its government operations under agentic AI management within two years, treating AI as an “executive partner” rather than a productivity tool. Agentic AI is categorically different from chatbots or copilots: these systems make decisions and take autonomous actions across multi-step workflows without human sign-off at each step. When a nation-state commits to this at scale, it sets a reference point that private sector organisations will increasingly be benchmarked against. For SME leaders, the implication is pointed: if a national government can run half its operations this way, the question of whether AI agents belong in your business has already been answered. The only remaining question is how far behind you are willing to be.
Meta Bets on Physical AGI
ROBOTS & PHYSICAL AI
Meta acquired Assured Robot Intelligence on May 1, a startup building foundation models that enable humanoid robots to understand and adapt to human behaviour in real time. The co-founders join Meta Superintelligence Labs with an explicit goal the company calls “physical AGI.” The strategy is the Android playbook: build the foundational software stack and license it openly to hardware manufacturers. Meta missed mobile, search, and cloud. With $115 billion committed to AI capital expenditure, it is not planning to miss humanoids. For SME leaders, this signals a compressed timeline to a world where robotic workers are a viable cost line. If the open-licensing play materialises, capability will reach businesses well below enterprise scale faster than most projections currently assume.
Grok 4.3 Halves the Price of Frontier AI
AI PRICING & MARKET
xAI released Grok 4.3 on April 30 at $1.25 per million input tokens: a reasoning model with always-on reasoning, multimodal input including video, and a one million token context window. It scores strongly on law and finance benchmarks, sitting just below Claude Opus 4.7 and GPT-5.5 on raw capability but shipping at a fraction of the cost. The model layer is commoditising faster than most predicted a year ago. For SMEs, the cost argument against deploying frontier AI across your operations is weakening quarter by quarter. The gap between businesses that have integrated AI deeply and those still evaluating it is now a cost-of-operations question, not just a capability one. The window for acting cheaply is still open. It will not stay open indefinitely.
OTHER MAJOR STORIES THIS WEEK
Claude Gets a Massive Compute Upgrade via SpaceX
INFRASTRUCTURE
Anthropic signed a partnership with SpaceX giving it full access to the Colossus 1 data centre, adding over 300 megawatts of capacity and more than 220,000 NVIDIA GPUs within the month. The immediate practical effect: doubled usage limits for Claude Code, removed peak-hour restrictions for Pro and Max subscribers, and substantially increased API limits for Claude Opus models. For businesses running Claude in production, this removes a meaningful bottleneck on reliability and output volumes. The partnership also includes exploration of orbital compute capacity, indicating the scale of ambition behind the infrastructure investment.
ServiceNow Launches an AI Workforce That Runs Your Business
ENTERPRISE AGENTS
At its Knowledge 2026 conference, ServiceNow unveiled its Autonomous Workforce: AI specialists that complete entire business processes end-to-end without human intervention across IT, HR, finance, legal, procurement, and security. ServiceNow is positioning itself as “the operating system of the AI-powered enterprise.” For organisations on the ServiceNow platform, this is an immediate operational question. For those not on it, this signals where enterprise software is heading: away from tools that assist humans at discrete tasks, towards agents that own entire workflow categories.
OpenAI Crosses $25 Billion in Revenue and Eyes a Public Listing
MARKET SIGNAL
OpenAI has surpassed $25 billion in annualised revenue and is in early discussions about a public listing, potentially in late 2026. Anthropic is approaching $19 billion. These are no longer startup-phase numbers. For SME leaders, the trajectory confirms these platforms are building durable, well-capitalised businesses with the resources to keep improving at pace. The tools your organisation uses today will be substantially more capable in twelve months, and the advantage of embedding them deeply now compounds.
QUICK DIGEST
Musk puts his own lab in fifth place, under oath.
INDUSTRY DRAMA
In testimony at an Oakland federal court, Elon Musk ranked Anthropic first, OpenAI second, Google third, Chinese open-source fourth, and his own xAI fifth. He also admitted xAI uses distillation from OpenAI models to train Grok, calling it “standard practice.” It was the first public, sworn confirmation of cross-lab distillation between US frontier AI companies. The admission tells you more about how these models are actually built than any press release.
The AI connectivity protocol hits 97 million installs
AI INFRASTRUCTURE
Anthropic’s Model Context Protocol crossed 97 million installs in March 2026. Every major AI provider now ships MCP-compatible tooling, and the Linux Foundation announced it will take MCP under open governance. MCP is becoming the standard layer for connecting AI models to business systems. If your organisation is planning any AI integration work, understanding MCP now avoids significant rework later.
GPT-5.5 instant is now the default ChatGPT model
MODEL RELEASE
OpenAI updated ChatGPT’s default model to GPT-5.5 Instant on May 5. It produces 52.5% fewer hallucinated claims on high-stakes prompts in medicine, law, and finance, and uses roughly 30% fewer words. Enhanced personalisation from past chats and connected Gmail rolls out to paid subscribers. If your teams use ChatGPT daily, their baseline just improved without any action on your part.
Musk receives a $158 billion pay package from Tesla shareholders
TECH & GOVERNANCE
More than 75% of Tesla shareholders approved Elon Musk’s 2025 compensation totalling $158 billion, one of the largest executive pay awards in corporate history. The approval rate is the more notable figure. Institutional investors, fully aware of Musk’s attention spread across Tesla, SpaceX, xAI, and DOGE, voted decisively in favour. A governance data point worth filing as a barometer of institutional confidence in frontier tech leadership.