Big week in AI and we are only halfway through it. Anthropic just blew past OpenAI on paper. China is locking down its AI talent like state secrets. And a study with a hundred thousand humans says the machines are now more creative than most of us.
Let me walk you through what landed and why I think it matters.
Anthropic Closes a $30B Round at a $900B Valuation
Anthropic just closed a funding round north of $30B at a pre-money valuation above $900B. Sequoia, Dragoneer, Altimeter, and Greenoaks each chipped in about $2B apiece.
That puts Claude's maker above OpenAI's $852B mark from March. First time Anthropic has been the most valuable private AI shop on the planet.
My take. Valuations at this altitude stop being about discounted cash flow and start being about who you want sitting at the table when the rules get written. Investors are not buying revenue here. They are buying a seat. Whether the business math ever catches up is a separate question, and one I would not bet the ranch on either way.
China Puts Travel Restrictions on AI Researchers
Beijing is now requiring government approval before top AI researchers at private firms like DeepSeek and Alibaba can travel abroad. The point is to keep the brains and the weights and the tricks of the trade inside the country.
This is the AI cold war moving from speeches into paperwork. Talent is the actual scarce resource in this game, not chips and not capital.
My take. When a government starts treating engineers like nuclear physicists, you know the stakes have changed. Expect the US to respond, probably not as bluntly, but it is coming. If you are an AI researcher with a Chinese passport and a US offer, the calculus just got a lot more personal.
Generative AI Beats the Average Human at Creativity
A study with about 100,000 participants found generative AI now scores higher than the average human on standard creativity tests. Not the top humans. The average.
That is a smaller flex than it sounds and also a bigger one than it sounds, depending on how you squint.
My take. Creativity tests are not the same as actually being creative. A divergent thinking score does not write a song that makes you cry. But the gap between machine output and what most working professionals do day to day is closing faster than people want to admit. If your job is to generate options, draft copy, or brainstorm angles, the floor is rising. Adapt or watch the floor rise without you.
WEF Says AI Will Drive Cyber Risk in 2026
The World Economic Forum put out a report saying 94 percent of orgs surveyed think AI is now the top driver of cyber risk. 87 percent flagged vulnerabilities in AI systems themselves as one of the fastest growing threats.
Translation. The thing your CISO is most worried about is the same thing your CEO wants to deploy everywhere by Q4.
My take. This is going to be a brutal year for anyone who shipped an AI feature without thinking hard about prompt injection, data exfiltration through tool calls, or agents with too much access. The wave of breaches tied to LLM tooling is coming. If you run an AI agent that can touch your inbox or your repo, audit the blast radius this week. Not next quarter. This week.