Meta wants permission to run your life. Or at least big parts of it.

The company just launched Muse, a personal AI agent designed to handle email, shopping, calendar management, and other daily tasks. It’s Meta’s latest attempt to close the gap with OpenAI, Anthropic, and Google in the broader AI race. But there’s a catch: Muse needs access to your email, payments, health services, and calendar to actually work.

That’s a big ask from a company that spent years rebuilding trust after Cambridge Analytica and countless other privacy scandals. Meta is billing Muse as a direct competitor to products like OpenClaw and Instinct, promising it can do everything from selling your car to booking plane tickets. The technology might be there. The question is whether anyone will actually hand over the keys.

This is Meta’s biggest consumer AI bet to date, and it hinges on something the company can’t engineer its way out of: whether people still trust Meta with their most sensitive data. The product launch represents a multi-billion-dollar strategy shift, one that acknowledges the company has fallen behind in AI and needs to catch up fast.

Meanwhile, Europe bets billions on homegrown AI

While Meta fights for consumer trust, Mistral just raised €3 billion at a €21 billion valuation. The French AI lab’s Series D was led by Samsung, Scaleup Europe, and PSG Equity, and it’s a clear signal that sovereign AI has become serious business.

Mistral’s funding round is part of a broader trend: governments and investors increasingly want AI models that aren’t controlled by American tech giants. Europe in particular has been pushing hard for domestic alternatives to OpenAI and Anthropic, driven by concerns over data sovereignty, regulatory control, and strategic independence.

The €21 billion valuation puts Mistral firmly in the top tier of AI companies, even if it’s still smaller than the American leaders. More importantly, it shows that “sovereign AI” isn’t just a talking point anymore. There’s real money flowing into companies that can offer governments and enterprises an alternative to handing their data over to U.S. firms.

The AI coding wars heat up

In another sign that AI investing remains frothy, Cognition just hit a $48 billion valuation. That’s higher than the multiple Cursor commanded before selling to SpaceX, and it suggests investors don’t think AI coding tools are headed for a winner-take-all outcome.

The Cognition valuation is notable because it flies in the face of conventional wisdom about developer tools. Historically, these markets consolidate quickly. But investors are betting that AI coding is different, that there’s room for multiple players serving different use cases, languages, and workflows.

Whether that’s true remains to be seen. But the fact that Cognition can raise at that valuation, even after Cursor’s high-profile exit, tells you where investor sentiment is right now.

Google Cloud makes an enterprise play

Google Cloud isn’t sitting still either. The company just announced a major partnership with Accenture, aimed at accelerating enterprise AI adoption. The deal focuses on forward-deployed engineers, people who can actually get AI systems up and running inside large organizations rather than just selling them.

It’s a tacit acknowledgment that deployment, not technology, is the current bottleneck for enterprise AI. Google has world-class models. What it needs is more boots on the ground to help companies actually use them. Accenture brings tens of thousands of consultants who already have relationships with the Fortune 500.

The partnership is Google’s latest move to catch up with Microsoft and Amazon in the cloud wars. All three companies are racing to prove they can turn AI hype into actual revenue, and enterprise deployment is where that battle will be won or lost.

The security response to AI accelerates

Both Microsoft and Google are changing how they ship software in response to AI-driven security threats. Microsoft just released patches for 972 vulnerabilities, 112 of them critical. Security teams warned the massive patch release is preparation for an expected wave of AI-assisted attacks.

Google, meanwhile, is moving Chrome to a two-week release cycle, doubling the pace of security updates. The company cited AI’s impact on the security landscape as a key driver for the change.

The moves are a reminder that AI isn’t just changing what software can do. It’s changing how fast software needs to evolve to stay secure. Defenders are speeding up their response times because they expect attackers to do the same.

Anthropic’s safety concerns go public

And finally, Anthropic is dealing with internal turbulence over AI safety. Jacob Coxon, a researcher who trained AI systems at the company, resigned over what he called a lax approach to safety. Hours later, another senior safety researcher at Anthropic said there’s more than a 10 percent chance AI “could kill all humans” by the end of the decade.

The public split is unusual for Anthropic, which has positioned itself as the safety-conscious alternative to OpenAI. Coxon’s resignation and the subsequent comments suggest that even at a company explicitly focused on AI alignment, there are sharp disagreements about whether the industry is moving too fast.

Whether you think the 10 percent estimate is credible or alarmist, the fact that senior researchers at leading AI labs are saying it publicly is worth noting. The companies building the most powerful AI systems are having very different internal conversations about risk than their public messaging would suggest.