Google's I/O keynote wrapped with over a dozen AI launches, Anthropic quietly overtook OpenAI in business adoption for the first time, and the EU drew a hard legal line against AI-generated deepfakes.
Google drops Gemini Omni and Gemini 3.5 Flash, and slashes AI Ultra to $100/month
At Google I/O 2026, Google released two new AI models: Gemini Omni, which can accept any input and generate video grounded in real-world knowledge, and Gemini 3.5 Flash, described as frontier intelligence combined with agentic action. Google also cut its AI Ultra subscription from $250 to $100 per month and replaced daily prompt limits with a compute-based model that refreshes every five hours. The Gemini app now has 900 million active users, up from 400 million at I/O 2025.
Why it matters: A price drop of 60% on the premium AI plan makes serious AI access much more affordable. The new compute-based usage model is also friendlier for heavy users who kept hitting daily caps.
Google announces Gemini Spark, Universal Cart, and AI-powered smart glasses
Alongside the model launches, Google announced a wave of consumer products: Gemini Spark (a personal AI agent inside the Gemini app), Universal Cart (an AI shopping assistant that works across stores), Ask YouTube, Gmail Live, and a new Daily Brief feature. Google also previewed smart glasses built with Warby Parker and Gentle Monster, powered by Gemini, planned for later this year. Google is processing 3.2 quadrillion tokens per month: up from 480 trillion at last year's I/O.
Why it matters: AI is no longer just a chatbot tab, it is being woven into shopping, email, and what you wear. The smart glasses in particular represent a push towards AI that is always present, not just available when you open an app.
Anthropic overtakes OpenAI in business adoption: Claude Code is the reason
For the first time, Anthropic's business AI adoption surpassed OpenAI's, reaching 34.4% of businesses tracked by the Ramp AI Index in May 2026. OpenAI's business adoption grew just 0.3% over the past year, while Anthropic quadrupled its share. The driving force is Claude Code, Anthropic's agentic coding tool, described as the fastest-growing product in the company's history. In separate news, OpenAI co-founder Andrej Karpathy announced he is joining Anthropic to work on pre-training research.
Why it matters: The AI market is not a fixed race between two giants. Tools that solve real workflow problems (like automating software development) are what companies are actually paying for. Karpathy moving to Anthropic is a significant talent signal.
EU AI Act adds two new hard prohibitions, including non-consensual deepfakes
On 7 May 2026, EU negotiators reached a provisional agreement amending the EU AI Act, the first changes since its adoption in June 2024. Two new AI uses are now outright prohibited: generating or manipulating non-consensual intimate imagery, and creating child sexual abuse material using AI. Both prohibitions take effect on 2 December 2026. The amendments also include timeline relief and targeted simplifications for businesses complying with other parts of the Act.
Why it matters: This is one of the clearest examples of regulation catching up with a specific harm, non-consensual deepfakes. The December deadline gives platforms a defined window to comply, not an open-ended aspiration.
Trump White House floats FDA-style AI regulation, then immediately walks it back
White House National Economic Council Director Kevin Hassett suggested the Trump administration was considering an executive order that would regulate AI models the way the FDA regulates drugs, requiring approval before deployment. Within hours, White House Chief of Staff Susie Wiles clarified that the administration was not in the business of "picking winners and losers," and a senior official told Politico the focus was on partnering with companies rather than regulating them. The mixed signals reflect genuine tension inside the administration about how to approach AI oversight.
Why it matters: Even an idea floated and walked back reveals that AI safety regulation is now a live topic at the highest levels of US government. Whatever direction this goes will affect what AI products can legally do in the world's largest market.
India's AI opportunity: $126B by 2030, and 90% of startups already using AI
The Bharat AI Startups Report 2026 estimates India's AI market at $126 billion by 2030, led by Enterprise AI (growing from $11B to $71B) and Consumer AI (scaling from $13B to $55B). India now has over two lakh registered startups, with nearly 90% using AI in some form. Over 600 startups are active in healthcare, agriculture, education, and manufacturing. The government's IndiaAI Mission is expanding access to compute and data, with the Union Budget committing ₹10,371 crore over five years to the programme.
Why it matters: These are not projections from an international consultancy, they reflect what is actually being built on the ground in India right now. For students and early-career professionals in India, the runway ahead is unusually long.
OpenAI and Anthropic are heading towards IPOs, but cheap AI from Chinese labs may rewrite their valuations first
Both OpenAI and Anthropic are moving toward public offerings, with OpenAI's confidential filing expected imminently. But a report from CNBC notes that Chinese AI labs are charging a fraction of what American labs charge for comparable capability, and a wave of cheaper Western alternatives is building fast. The central premise of their valuations (that premium models command premium prices) may be under pressure before the prospectuses even land. Watch how both companies position their enterprise pricing and product differentiation in the coming weeks.
Cursor ships a Claude Opus 4.7 rival at a fraction of the cost • Amazon Alexa gets AI podcasts • OpenAI launches its Deployment Company
Cursor, the AI coding tool, released a new coding model it says competes directly with Claude Opus 4.7 at significantly lower cost: another data point in the race to build capable-but-cheap alternatives. Amazon's Alexa is now generating personalised AI podcasts on demand. And OpenAI launched the OpenAI Deployment Company, backed by an acquisition of Tomoro, aimed at speeding up enterprise AI rollouts.