UK's CMA wants AI assistants on Google's choice screens
Regulation caught up with the chatbot-as-search-engine era today, OpenAI put mathematicians on the payroll of its math model's aftermath, and an agent startup says big SaaS is becoming "a database." Here's the day.
The UK's competition watchdog has proposed forcing Google to offer Android and Chrome users a choice of search engine when they first set up a device — and to repeat the question once a year. The CMA's updated proposals go further than choice screens for Bing and DuckDuckGo: AI assistants such as ChatGPT and Perplexity would be allowed onto the screens too, provided they meet the regulator's technical and security criteria — a direct answer to people increasingly "searching" by asking a chatbot instead of a search box. CMA chief executive Sarah Cardell said the regulator is "responding quickly to the rise of AI assistants, which people are increasingly using as their search service, but aren't included on Google's choice screens." The proposals would also require featured providers to fairly attribute publisher content, so users can see where an answer came from — a small clause with large consequences for the AI-search traffic war. Google has operated under the UK's strategic market status regime since October 2025, and the CMA has been tightening the screws since June; this consultation closes October 9, with a final decision on the conduct requirements expected by year-end. The takeaway: the choice screen is quietly becoming the regulatory beachhead for AI assistants — whoever gets listed there gets distribution no app store battle can match.
OpenAI says an internal model has resolved more than 100 open math problems since it began training on August 28 — and it has now handed nine leading mathematicians an advisory role in how those results reach the field. The group, hosted at Princeton's Institute for Advanced Study, includes Timothy Gowers, Martin Hairer, and Melanie Matchett Wood, and it can weigh in on vetting, releasing, and crediting results. It is not paid by OpenAI and has no say over how fast the model produces new results — which is the part the field keeps objecting to. We covered that complaint when 27 Fields Medalists signed a letter accusing labs of rushing results out — Mathematicians ask whether OpenAI can be trusted with unpublished math. Today's announcement reads as OpenAI's answer to that letter: humans get a formal seat in the loop, but the loop keeps running at model speed.
Agent startup Ema has raised $77 million in a Series B led by Bengaluru's Creaegis, with Accel, Section 32, and Prosus increasing their stakes, as AI begins competing directly for enterprise software and IT services budgets. According to TechCrunch, the round more than quadruples Ema's 2024 valuation and takes its total funding to $140 million. The company coordinates fleets of AI agents that run multi-step workflows across HR, IT, and finance on top of the apps a company already owns, and cofounder Surojit Chatterjee's pitch is blunt: big SaaS platforms are "mostly becoming like a database" that agents wrap around and, in some cases, replace outright. Ema claims more than 50 active enterprise deals, over 1 million active users, roughly 180% net dollar retention, and about 80% gross margins — figures that are the company's own and not yet independently confirmed, and the exact valuation went undisclosed.
What to watch: the CMA consultation closes October 9 — if the final requirements name AI assistants explicitly, expect Brussels to start drafting its own version within weeks.
Given a first-run choice screen, would you set an AI assistant as your default search? Tell us in the comments.
Sources: UK CMA (gov.uk) · Engadget · OpenAI · The Rundown AI · TechCrunch · Moneycontrol
