CHATGPT JUST LOST ITS MAJORITY. THE AI RACE IS FINALLY ON.
So here is something that nobody thought would happen so fast. ChatGPT, the product that basically invented the mass-market AI assistant category and had been sitting fat and comfortable at over half of all AI assistant usage worldwide, has officially dropped below 50% market share for the first time. According to analytics firm Sensor Tower’s 2026 State of AI Report, OpenAI’s flagship still has a staggering 1.1 billion monthly users. But the rest of the field has caught up enough that ChatGPT no longer commands a majority. Google’s Gemini is sitting at 662 million monthly users. Anthropic’s Claude is at 245 million. And then you have Grok from xAI quietly picking up everyone who followed Elon Musk into whatever that is.
What’s interesting here is not just the numbers, it’s what’s moving them. People are actually switching between AI assistants now, which was not happening a year ago. You didn’t just pick one and stick with it out of loyalty or laziness the way you did with, say, your email app. Something specific triggered it. The biggest single event that caused ChatGPT uninstalls, according to Sensor Tower, was OpenAI’s February deal with the US Department of Defense. A significant chunk of users decided they did not want their AI assistant also moonlighting for the Pentagon, and they went and downloaded Claude or Gemini instead.
There is something almost poetic about that. OpenAI spent years telling everyone it was building artificial general intelligence for the benefit of all humanity. The moment it signed a DoD contract, hundreds of thousands of people said, “You know what, I’m good,” and quietly moved on. Users apparently draw a line somewhere, and that line turns out to be somewhere around weapons procurement.
The bigger picture here is that the AI assistant market has entered a new phase. The land grab is over. The moat is not as deep as OpenAI thought it was. ChatGPT’s brand is still dominant, and 1.1 billion users is not exactly a crisis, but the era where you could just sit back and coast on first-mover advantage appears to be ending. Gemini has Google’s entire search empire pushing people toward it. Claude has the reputation of being the “responsible” one, which ironically got a boost when the government tried to ban its most powerful models and everyone immediately wanted them. Grok has whatever Musk’s 200 million Twitter followers do when he tells them to download something.
What nobody is really talking about is that all of this user switching is happening before the AI assistants have actually gotten dramatically better at anything. The next wave of improvements, when they come, is going to reshuffle things again. The company that figures out how to be genuinely useful for everyday tasks, not just impressive in demos, is going to win the second round. Right now everybody is still fighting over who has the most impressive benchmarks. The user who just wants to split a dinner bill or draft an email does not care about benchmarks. They care about the thing working correctly the first time.
The monopoly is cracking. This is worth paying attention to.
APPLE IS PUTTING AI IN EVERYTHING ON YOUR IPHONE AND MOST PEOPLE HAVEN’T NOTICED YET
Everyone spent the last two weeks talking about the new Siri because that was the headline at WWDC, and yes, a Siri that finally understands context and can actually do things is a big deal. But while people were arguing about whether Apple was too late to the AI party, the company was quietly stuffing about a dozen genuinely useful AI features into iOS 27 that nobody seems to be talking about. And some of them are actually good.
Take the bill-splitting feature. You take a photo of the restaurant receipt, Apple Intelligence reads it, breaks down who ordered what, calculates the tax and tip per person, and lets you send individual payment requests directly through iMessage and Apple Cash. This sounds small. It is not small. Anyone who has stood outside a restaurant at 11pm watching six adults argue over Venmo requests knows this is the kind of thing that genuinely improves your social life.
Then there is the password updater. Your password manager already flags compromised passwords. The new thing is that Apple Intelligence will now go to those websites on your behalf, log in, and change the password automatically. You do not do anything. It just handles it. This is agentic AI doing something real in your life, not in a demo, not in a press release. Your actual passwords, getting actually updated, without you having to click through twelve menus.
There is also a feature called Call Context, which surfaces relevant information on your screen when you call a company’s customer service number. Calling about your flight? Your confirmation code appears on the call screen, pulled from your email, processed entirely on the device. Nobody sees it but you. This is the kind of thing that makes you wonder why it took this long to exist.
And then there is the Shortcuts update, which lets you describe what you want your phone to do in plain English and it just builds the automation. Apple is calling it the AI version of Shortcuts but it is basically vibe coding for non-programmers, which means the most powerful automation tool on iPhone is finally accessible to the 95% of people who never learned how to use it.
Apple’s strategy here is interesting. Rather than bet everything on a single AI assistant experience that users have to consciously adopt, they are stitching intelligence into the apps and services people already use every day. The AI is in the background. It is doing things without being asked. And because it is all processed on-device for most features, they get to keep making the privacy argument. Quiet, practical, embedded. That has always been Apple’s play, and it might actually work better than building the world’s most powerful chatbot and hoping people remember to open it.
WASHINGTON CLAIMS CHINA MIGHT HAVE THE WORLD’S MOST EXPENSIVE AND DANGEROUS CHIPMAKING MACHINE. ASML SAYS THAT IS IMPOSSIBLE. SOMEBODY IS LYING.
Source: TechCrunch | Bloomberg
This one is wild. Commerce Secretary Howard Lutnick, who you might remember from the Anthropic crackdown last week, apparently called ASML’s top executives and told them he was concerned that one of their extreme ultraviolet lithography machines, an EUV machine, might have ended up in China. ASML makes the only EUV machines that exist on Earth. They cost somewhere between $200 million and $380 million each depending on the model. They are arguably the most complex pieces of machinery ever built by human beings. And they have been banned from export to China since the first Trump administration.
ASML said, essentially, “No. That is not possible. We track these machines. None of our EUV systems are in China.” The US government said it had evidence. Then the US government declined to actually show the evidence to ASML, or to Bloomberg, or apparently to anyone. Senior administration officials told Bloomberg they have evidence that ASML shipped EUV-related components and transport equipment to China. Not an entire machine, mind you. Components and transport equipment. Which is a very different claim than “China has a full EUV machine.”
Here is what makes this significant beyond the he-said-she-said of it. EUV machines are the only way to manufacture the most advanced chips in the world. If China somehow had access to one, even a partial one, it would represent a potentially catastrophic failure of the entire Western semiconductor containment strategy. The whole point of the chip export controls, the restrictions on ASML shipping to China, all of it, was to prevent China from being able to manufacture chips at the frontier. A single EUV machine would not solve China’s chip problem overnight. But it would be a significant intelligence failure and would require a major response.
The fact that the US government is making this claim to ASML but then refusing to provide the evidence to either ASML or journalists is strange. Either they have compelling evidence and are protecting sources and methods, in which case you would expect them to at least share it privately with ASML under NDA, or the evidence is thinner than implied and this is another case of the Lutnick Commerce Department using allegations as leverage. Given what happened with Anthropic the previous week, the pattern is at least worth noticing. Howard Lutnick has been very busy making dramatic calls to tech companies with dramatic claims, and then the specifics turn out to be complicated. That does not mean he is wrong about ASML. It just means you should not immediately assume he is right either.
POLL SAYS ONLY 16 PERCENT OF AMERICANS THINK AI WILL BE GOOD FOR SOCIETY. THE INDUSTRY’S PR PROBLEM IS GETTING WORSE.
You can spend as much time as you want reading about GPT-5.5 benchmarks and billion-dollar funding rounds and Jensen Huang on stage in a leather jacket. But here is the number that should make every AI executive uncomfortable: only 16 percent of Americans believe artificial intelligence will have a positive impact on society. That is a new study, and it is not a number that an industry can dismiss as technophobe noise from people who do not understand the product.
Think about what 16 percent means in context. That is roughly the same approval rating as Congress on a good day. It means that if you are in a room of ten people, statistically one person and maybe part of a second person thinks AI is going to be good for the world. The other eight and a half people are somewhere between skeptical and actively worried. And these are not Luddites who refuse to use technology. These are people with smartphones, streaming subscriptions, and GPS navigation. They use tech constantly. They just think this particular wave of tech might be different, and not in a good way.
The AI industry has spent three years telling us that this technology will cure cancer, accelerate scientific discovery, make everyone more productive, and generally save civilization. And the public has watched Goldman Sachs announce it is eliminating jobs, watched Microsoft and Google lay off thousands while reporting record profits from AI investment, watched their friends and colleagues get automated out of work, and read headlines about AI being used for surveillance and deepfake fraud and autonomous weapons. The gap between the promotional narrative and the lived experience has become enormous.
This is not a problem you solve with a better press release or another TED Talk about AI for good. The public is doing a simple calculation: who is benefiting from this, and is it me? Right now the answer for most people is “the shareholders are benefiting and it is not me.” Until that changes, these numbers are not going to change. The 16 percent figure is a warning sign, and the industry would do well to take it seriously rather than writing it off as fear of change. Fear of change and fear of getting replaced are two very different things.