Only 15% of Americans trust AI companies to make decisions about AI. That puts the people building the technology below the federal government. The month had more where that came from. A government order took two of Anthropic's models offline within 72 hours of launch. Anthropic and OpenAI both filed for IPO. Apple rebuilt Siri on a competitor's AI. And a former OpenAI executive resurfaced with a $2 billion lab.

You don't have to follow AI news to feel its effects. These seven stories broke in June while most founders were heads-down in their own businesses. Each one changes something about what you build, what you sell, or how dependent you are on the companies behind your tools. Here is what happened and what to do with it.

The biggest AI stories of June, starting with trust in AI companies

Only 15% of Americans trust AI companies to make decisions about AI

Anthropic's first Public Record survey of nearly 52,000 Americans found that only 15% trust AI companies to make decisions about how AI is developed and used. That was the lowest figure for any institution tested, below the federal government at 20%, state and local government at 19%, and international bodies at 20%. Separately, 64% named AI-driven job loss as their top fear, and that held across every state and every party.

The counterintuitive part is who feels it. Americans who use AI every day at work are 16 percentage points less worried about losing their job than those who never touch it (54% against 70%). The people most afraid of AI are the ones not using it. The answer is hands-on practice. Get building with it this week. Once you understand the tool, you fear it less.

A government order pulled two of Anthropic's models offline in 72 hours

On June 12 the US government issued an export control directive requiring Anthropic to suspend all access to Claude Fable 5 and Claude Mythos 5, its two most powerful models, for any foreign national inside or outside the US, including its own foreign national staff. The models had launched three days earlier. Anthropic cannot filter users by nationality in live time, so it took both offline entirely to comply. Every other Claude model stayed up. The stated reason was national security: the government believed a jailbreak method had been found that could enable dangerous use of Fable 5.

Frontier AI access is now a geopolitical asset a government can switch off overnight. If your business runs on one model from one provider, you have a single point of failure that has nothing to do with the quality of your work. Keep your prompts, your workflows, and your trained expertise portable. Build on your own IP so that when a tool goes dark, your business does not go with it.

Anthropic and OpenAI both filed for IPO in the same week

Anthropic submitted a confidential draft S-1 to the SEC on June 1, targeting a valuation around $965 billion on a reported $47 billion revenue run rate. OpenAI followed ten days later, targeting roughly $1 trillion with Goldman Sachs and Morgan Stanley running the offering. OpenAI announced its own filing, saying "we expect it to leak so we're just announcing it." Together the two companies represent nearly $2 trillion of expected public-market value.

When the two labs that set the direction of AI both move to public markets in one week, the industry is institutionalising. Public companies answer to shareholders, and that changes product priorities and pricing on a quarterly clock. The more your business runs on your own audience and your own trained AI, the less those pressures reach you.

Apple rebuilt Siri on Google Gemini in Tim Cook's last keynote

At WWDC on June 8, Apple unveiled a rebuilt Siri running on a custom Google Gemini model, licensed for a reported $1 billion a year. Siri is now a standalone app with screen awareness that can act across other apps. Simple tasks run on the device, moderate ones go to Apple's Private Cloud Compute, and the heaviest reasoning routes to Google Cloud on Nvidia chips. The keynote was Tim Cook's final one as CEO before he moves to executive chairman on September 1.

Apple spent years insisting it would build AI on its own terms, then licensed a competitor's model for a billion a year. Frontier capability has become a procurement decision, even for the most valuable company in the world. The assistant in your clients' pockets became far more capable overnight. The model underneath is a commodity. What sits on top of it, your expertise and your point of view, is the part nobody can license.

An AI CEO asked governments for the power to block AI models

On June 10, Anthropic CEO Dario Amodei published " Policy on the AI Exponential ," calling for mandatory third-party safety testing and legal authority for governments to block or reverse frontier model deployments that fail. He proposed FAA-style pre-approval before major models ship, covering safety, economics, science, civil liberties, and geopolitics. The next day Anthropic announced Claude Corps , a $150 million commitment to place 1,000 AI fellows inside US nonprofits, each paid $85,000 a year for twelve months.

Read the two announcements as one argument. The head of one of the most powerful AI labs is saying that the disruption is coming and that the builders owe something to the people it displaces. For anyone helping clients through AI changes, that is a clear, simple thing to tell them. Use it. And the company is backing the claim with $150 million. The people causing the disruption are starting to fund the people it hits.

The AI subscription price war started at $4.99 a month

On June 9, Google cut its AI Plus subscription from $7.99 to $4.99 a month and doubled storage from 200GB to 400GB. OpenAI is reportedly weighing cuts of its own. Meta went the other way, launching its first paid AI subscriptions between $7.99 and $19.99. Subscriber growth has slowed across the sector and enterprise buyers are spending more carefully, so the labs that spent two years competing on model performance are now competing on price.

Five dollars a month removes the last excuse not to experiment. As the subscription layer becomes interchangeable, the thing that separates AI tools is the quality of what they are trained on. A coach or consultant who has built an AI on their own expertise is working from a different foundation than one running a generic model anyone can rent for pocket change.

Mira Murati resurfaced with a $2 billion lab chasing instant AI

After eighteen months of near-silence, former OpenAI CTO Mira Murati appeared at Bloomberg Tech on June 4 to describe what Thinking Machines Lab is building : AI that responds at the speed of conversation, targeting sub-200-millisecond latency for continuous human-AI dialogue. The company has raised $2 billion at a $10 billion valuation. Its first product, Tinker, is an API for fine-tuning open-source models. Murati spoke of $50 billion in longer-term ambitions but said the priority now is shipping.

Murati ran the launches that defined ChatGPT. Her choice to build around speed of response rather than smarter reasoning points to where AI experiences are heading: interaction fast enough to feel like talking to a person, not querying a search engine. If you are thinking about how AI products will feel in two years, that is the direction. The people who built the first generation of AI are building the next one with a clear view of what the first one missed.

Why so few Americans trust AI companies, and what to do about it

This is a good month to be a founder paying attention. A government can take a model offline. Two labs are about to enter public markets. Apple rents its AI from a rival. The CEO of a leading lab wants laws to restrain his own industry. Underneath the trust gap sits one usable fact: the people most afraid of AI are the ones not using it, and the people building on their own expertise are the least exposed to any single provider's decision. Own your IP. Stay portable. Get hands-on now. The people closest to AI are desperately telling you to take action.

Get my free playbook on building your business with AI.