18 May 2026: Apple’s Siri Privacy Bet and AI’s $80 Billion Revenue Mountain (AM)

This morning brings two stories that capture where AI sits in May 2026. Apple is preparing a privacy-first Siri overhaul that frames data control as a competitive feature, while fresh figures from The Information show Anthropic and OpenAI now hoover up nearly nine in every ten pounds spent on AI startup products. The shape of the industry is hardening into a duopoly even as the consumer pitch shifts.

Apple is selling privacy as a Siri feature ahead of June’s WWDC. A run of reports yesterday, led by Bloomberg’s Mark Gurman and matched by 9to5Mac, Engadget and TechCrunch, says the upcoming standalone Siri app will let users set chats to auto-delete after 30 days, after a year, or keep them indefinitely. The retention panel will sit inside Siri’s own settings rather than in the system-wide privacy menu, which suggests Apple wants the choice to feel like a product feature rather than a buried preference.

Apple’s pitch is a direct contrast to ChatGPT and Gemini, which only offer temporary incognito sessions. The wrinkle is that the new Siri will reportedly run on Google’s Gemini models, processed on Apple’s private cloud compute infrastructure rather than handed straight to Google’s servers. The app is expected to ship with iOS 27 and to keep its “beta” label at launch, repeating the cautious roll-out pattern Apple used with the original Apple Intelligence in 2024.

For UK readers, the practical question is what changes on day one. If you already use ChatGPT or Claude, an Apple-built alternative on iPhone with default deletion may matter more than a model benchmark. Anyone running a small business on iOS should also note that Siri will sit inside the apps it integrates with, so data retention settings will affect calendar, mail and contact information at once. Our guide to AI tools worth paying for right now covers how to think about that trade-off.

Anthropic and OpenAI now capture 89 percent of revenue from the top 34 AI startups. The Information published an analysis yesterday afternoon showing the group is generating roughly $80 billion (about £63 billion) in annualised revenue, up 112 percent in six months. The two leaders account for almost all of that figure, with the remaining 32 firms splitting the residual $9 billion between them.

The concentration is striking even by the standards of a winner-takes-most market. Anthropic disclosed in its May 6 compute update that its run-rate revenue had passed $30 billion (about £24 billion), driven heavily by Claude Code adoption and enterprise rollouts. OpenAI’s product mix shifted in March, when its applications division consolidated under Fidji Simo and announced the desktop super app merging ChatGPT, Codex and the Atlas browser.

For consumers and small businesses, the takeaway is pragmatic. When picking AI tools, the two biggest providers are also where product investment is concentrated, so support and feature velocity will likely favour them. Customers of smaller specialised AI startups should ask harder questions about runway, since the gap between leaders and the rest is widening, not narrowing.

Three AI startups landed combined funding of $95 million in a single day. Sales automation startup Monaco raised a $50 million Series B led by Benchmark, taking total funding to $85 million. AI government permitting platform GovWell raised a $25 million Series A from Insight Partners. Back-office automation startup Ciridae raised a $20 million seed from Accel with a16z participating. All three deals were reported by Business Insider, Axios and Fortune yesterday.

What links them is the workflow focus. Each picks a process that small and mid-sized organisations actually spend money on, rather than chasing a model breakthrough. GovWell tackles licensing and permitting paperwork. Monaco automates sales outreach. Ciridae targets internal operations for “real-economy” businesses, a category that covers manufacturers, distributors and trades. Capital is flowing toward AI applied to specific industries, not at general intelligence, and the cheque sizes show it.

UK small businesses watching the funding flow should treat it as a signal of where buyers will land in 12 months. A useful precaution is to set out your own rules first: our simple AI policy guide for small businesses covers what to write down before adopting any new tool.

The Musk versus OpenAI trial closes with trust as the central question. TechCrunch reported yesterday that the final days of the trial returned repeatedly to whether OpenAI chief executive Sam Altman is trustworthy. The case, which Elon Musk filed contesting OpenAI’s transition from non-profit to capped-profit, is now in the hands of the court and a verdict is expected in the coming weeks.

The legal mechanics matter less to consumers than what the answer signals. If the court finds OpenAI’s governance changes lawful, the structure that has powered the company’s commercial expansion stands intact. If it rules against OpenAI, fresh structural questions could ripple into recent funding rounds and the rollout of the desktop super app. Either way, watch for procedural detail in the next 7 days.

Here is everything else worth knowing from this morning’s AI news.

• Apple’s Siri will run on Google Gemini: The revamped app will use Google’s models processed on Apple’s private cloud, not on Google servers.
• PwC deploys Claude across enterprise functions: The consulting firm will use the assistant for technology builds, deals and reinvention work for clients. [14 May]
• Anthropic and the Gates Foundation form $200 million partnership: The deal targets global health and development work, the largest grant-linked AI partnership to date. [14 May]
• Claude for Small Business launches: A new tier targets firms with fewer than 50 employees, sitting below the existing Team plan. [13 May]
• Microsoft AI chief Mustafa Suleyman gives 18 months: He told Fortune AI will automate most white-collar work within that window. [May]
• Pew and Gallup data: most Americans do not trust AI: Or the people running it, a Verge analysis of the surveys finds. [17 May]
• Show HN: Semble, code search for agents: A new open-source tool claims 98 percent fewer tokens than grep for agent code search. [17 May]
• 34 AI startups now generate $80 billion annualised: Up 112 percent in six months, with Anthropic and OpenAI taking 89 percent. [17 May]

The next data point worth marking is Apple’s WWDC keynote in early June. If Siri ships with auto-delete enabled as the default, expect Google and OpenAI to follow within weeks. If retention is set to indefinite by default, it tells you the inside-the-house tension over training data beat privacy as a competitive lever.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s daily guide to developments in artificial intelligence, published every morning.