This week the AI industry moved on two fronts simultaneously: the business model and the body count. ChatGPT switched on pay-per-click advertising, a humanoid robot demolished the human half-marathon record, and Anthropic asked users for their passports. Meanwhile, Meta announced it is trimming tens of thousands of jobs to fund AI, and OpenAI quietly crossed $25 billion in annual revenue. The pace is not slowing.
THIS WEEK’S TOP 3 MUST-KNOW CONCEPTS
ChatGPT Just Became an Ad Platform
BIG TECH
OpenAI has activated cost-per-click advertising inside ChatGPT, letting brands bid $3-$5 per click directly alongside AI responses. This is not a trial: internal forecasts project $2.4 billion in ad revenue for 2026 alone, with $102 billion targeted by 2030. For SME leaders, this is a double signal. First, the tool your staff uses daily is about to show them ads, which means your competitors can now target your team inside their AI assistant. Second, if you advertise, ChatGPT is a new channel worth evaluating before the cost of entry rises. Early adopters on Google and Facebook locked in audiences at a fraction of today’s rates. The window is open.
A Robot Just Beat Every Human at Running
ROBOTICS
Honor’s Lightning humanoid robot completed the Beijing half-marathon in 50 minutes 26 seconds, running fully autonomously past 12,000 human competitors. The human world record is 57:20. Last year’s winning robot needed 2 hours 40 minutes to cover the same distance. In under 12 months, autonomous robot performance on a physically demanding real-world course has improved by roughly 70%. For SME owners in logistics, warehousing, manufacturing, or any sector with repetitive physical tasks, this is your clearest signal yet that robotic labour is not a distant prospect. Start identifying the roles in your business that are structured, repetitive, and physically defined.
Anthropic Wants Your Passport to Use Claude
AI ACCESS
Anthropic quietly introduced government ID verification for select Claude users, becoming the first major AI chatbot provider to require a physical passport, driver’s licence, or national ID card alongside a live selfie scan. The rollout targets higher-tier plans and advanced features, not all users. Neither OpenAI nor Google currently requires this. Reaction has been largely negative, with critics pointing out that Anthropic built its brand on privacy. The practical implication for SMEs: if your team relies on Claude for sensitive or advanced work, factor this friction into your AI tool stack decisions. More broadly, identity layers on AI platforms are coming across the industry. It is worth reviewing which AI tools your staff use and whether your data governance policies cover them.
OTHER MAJOR STORIES THIS WEEK
Meta Planning to Cut Up to 20% of Its Workforce
AI ECONOMY
Meta is preparing to slash roughly 10% of its global workforce in an initial round of cuts expected around 20 May 2026, with additional, potentially deeper cuts planned for the second half of the year. Together, the rounds could affect up to 20% of Meta’s roughly 80,000 employees. The publicly stated driver is offsetting rising AI infrastructure costs. For SME leaders, this is a concrete illustration of AI’s economic mechanics at scale: even a company generating tens of billions in profit is restructuring headcount to fund AI investment. The question your board should be asking is not whether AI will affect your workforce, but how deliberately you are planning for that conversation.
Perplexity Ships Personal Computer: A Physical AI Brain for Your Desk
AI AGENTS
Perplexity has released Personal Computer, a Mac mini running its proprietary AI OS that gives the cloud-based Perplexity agent persistent local access to your files, apps, calendar, and sessions, enabling it to run proactive tasks around the clock without cloud round-trips. Combined with Perplexity’s existing Computer agent, which executes multi-step autonomous workflows for Perplexity Max subscribers, this creates a desktop AI employee that is always on and always context-aware. For SMEs evaluating agentic AI, this is the most consumer-accessible form factor yet. It is not ready for enterprise procurement, but it signals where the market is heading: a dedicated, always-on AI layer sitting inside your office infrastructure.
OpenAI Launches Images 2.0 with Reasoning and Near-Perfect Text
TOOLS
OpenAI has released ChatGPT Images 2.0 alongside Images Thinking, which applies reasoning to image generation for the first time. The headline improvements are near-perfect text rendering in generated images, elimination of the colour distortion that plagued the previous generation, and the ability to generate UI mockups indistinguishable from real screenshots. Images Thinking is available to all paid ChatGPT plans. For SMEs, this significantly lowers the barrier to producing polished marketing visuals, product mockups, and presentation graphics without a designer. If your team is not yet experimenting with AI-generated imagery for day-to-day content, the quality threshold has just cleared the bar for practical business use.
QUICK DIGEST
Anthropic Launches Claude Design
TOOLS
Anthropic has released Claude Design, competing directly with Figma for design and prototyping work. Initial reviews are mixed: strong benchmarks, rough edges in workflow integration. For SMEs paying for design tools or freelancers, it is worth testing before your next renewal.
OpenAI Passes $25 Billion Revenue, Eyes IPO
AI BUSINESS
OpenAI has surpassed $25 billion in annualised revenue and is taking early steps toward a public listing, potentially late 2026. Anthropic is approaching $19 billion. These are current run rates, not projections. For SMEs, your AI tool vendors are here to stay, but pricing will shift as they build toward profitability.
DeepMind Hires Philosopher for AGI Questions
AI ETHICS
Google DeepMind has brought on philosopher Henry Shevlin to explore AGI and machine consciousness. A frontier lab employing a philosopher of mind, rather than just engineers, signals that the hard questions are no longer theoretical. SME leaders do not need answers, but governance policies for AI deployment should account for this uncertainty.
Grok 5 Coming in Q2 at 6 Trillion Parameters
AI MODELS
xAI is expected to release Grok 5 in Q2 2026, built on a 6-trillion-parameter architecture. Whether that scale translates to practical gains is unproven. For SMEs evaluating AI tools, the message is simple: the gap between what models can do and what most businesses actually use them for remains enormous. Focus on deployment, not benchmarks.