Author: Sharilyn Deseo

A proposed AI sovereign wealth fund in the United States would open up a striking possibility: you could own a slice of the AI boom through public stakes, and might even get annual payments or other broader benefits from it.

It’s easy to see why the idea is catching on. In market forecasts that get cited all the time, the US AI market grows from $173.56 billion in 2025 to $976.23 billion by 2035. The global AI market, meanwhile, goes from $130 billion in 2023 to nearly $1.9 trillion by 2030. And by many measures, wealth has risen 70% since late 2022, with the top 1% taking a larger share.

But once you get past the slogan, the mechanics get messy. Senator Bernie Sanders’ version would use a one-time 50% stock tax on large AI companies to build a $7 trillion fund. After that, you still have to sort out which firms actually qualify, what their share values are, how public ownership would function, who would run the fund, what it would invest in, and how any payouts would be handed out.

Then come the governance risks. Would the US government hold the shares, vote them, or trade them? How would it avoid conflicts and political pressure? And would public ownership end up distorting competition, slowing innovation, or pushing Washington into the role of picking winners and losers?

If you like the idea, this is a proposal worth watching. Even so, analysts generally lean toward taxes, antitrust , or direct spending instead, even as South Korea explores a version of the model that the United States would be trying on a larger, more contentious scale.

For now, you can’t opt into a US AI sovereign wealth fund.

Two years into the push to use artificial intelligence to answer customers faster, speed up internal work, and trim costs, a lot of artificial intelligence strategies have run into the same wall: trust. Reliability, transparency, and accountability aren’t side issues anymore. They’re the issue.

The gap is getting harder to ignore. Executives celebrate the speed; users want something else. They want clear reasoning, some way to challenge a wrong answer, and a person who’s accountable when something breaks. A chatbot or workplace assistant may look efficient in a report, but when it’s opaque, inconsistent, or confidently wrong, it chips away at confidence. That helps explain why trust in artificial intelligence is still low around the world, and why so many projects stall in pilot mode or never produce clear returns.

The deployments that are holding up better treat this as a trust design problem. They’re building systems that behave predictably, offer explanations people can follow, and keep human oversight visible, so output can be reviewed, edited, or overridden. That matters most because persuasive answers that are still inaccurate are common.

If you’re rolling out artificial intelligence across support tools, workplace assistants, or newer agentic systems, keep an eye on this. Shadow tools, sensitive data exposure, and decisions that are hard to trace are pushing organizations toward stronger governance, models people can actually explain, better training, and clearer human accountability.

In Q1 2026, face-licensing platforms in China are spreading alongside the country’s microdrama boom. The model looks a lot like a stock-photo marketplace: you upload a handful of images, sign a consent form, and let producers rent your likeness for speaking parts, animated appearances, and recurring roles in ads and short dramas.

CapCut Download

ActID and New Claw pitch this as a way to make casting faster and cheaper for creators, while giving ordinary people a bit of side income. The China Netcasting Services Association says generative video showed up in more than 95% of the 128,000 microdramas released in Q1 2026. Reported contributor payouts range from about $15 to $700. On ActID, creator fees run from 99 to 500 yuan per episode, with the platform taking a 10% cut. On New Claw, minimum pricing starts at 500 yuan per image. Some reports also say a 60-episode synthetic drama can be produced by a small team in roughly 20 days.

If you follow AI video, keep an eye on this.

China already requires consent for facial data and bans unauthorized fabrication. But once those files leave a platform, the rules around duration, modification, misuse, and reuse still vary from one set of terms to another. The Guangzhou Internet Court has reportedly dealt with about 700 generative-media disputes over the past three years, according to reports.

ActID and New Claw are available in China, where draft Cyberspace Administration of China rules from April 2026, along with labeling rules that took effect in September 2025, are putting tighter controls in place. All of that is happening while actors deal with lower pay, fewer jobs, face-cloning risks, and contracts that fold likeness rights into the bundle.

Genesis AI is back in fundraising talks, and the figure being discussed is large: about $500 million at a pre-money valuation of roughly $3 billion, according to a Bloomberg report published today.

ChatGPT Download

If a round like that comes together, it would be a major leap for Genesis AI, a company most people had barely heard of until it emerged from stealth in July 2025. That launch came with a $105 million seed round co-led by Eclipse Ventures and Khosla Ventures.

Nothing is locked in yet. Bloomberg says the deal isn’t final, and the terms could still shift. Even so, if Genesis AI lands anywhere near that level, it would rank among the biggest follow-on jumps in the race to build a shared robot “brain.”

The company’s pitch is built around GENE-26.5, a universal model designed to work across different robot form factors. That includes the Eno wheeled robot, dexterous hands, an in-house physics engine that can generate synthetic data hundreds of thousands of times faster than real time, and a $300 sensor glove for capturing human motion.

The idea is to get robots trained in simulation to perform in the real world. Genesis AI says a single shared model can handle jobs as different as cooking, solving a Rubik’s Cube, and playing the piano.

If you follow general-purpose robotics, this one is worth keeping an eye on. The rival list is already crowded: NEURA Robotics, Humanoid, Standard Bots, Skild AI, Physical Intelligence, Field AI, and Scout AI are all in the mix. Genesis AI was founded by Zhou Xian and Theophile Gervet across Paris and the Bay Area, and its backers include Eric Schmidt, Xavier Niel, Bpifrance, and HSG. Bloomberg also says Premji Invest may lead the round, with HSG and Northzone considering putting in more money.

Chinese AI company Z.ai, the outfit formerly called Zhipu, is getting close to the $1 billion annual sales line in 2026. Bloomberg says the company is reaching that level through a combination of booked revenue and annualized recurring revenue.

ChatGPT Download

Bloomberg reports that Z.ai brought in 724 million yuan in 2025, roughly $100 million, up 131.9% from a year earlier. JPMorgan projects about 4.6 billion yuan in 2026 and 30.9 billion yuan by 2028.

Taken together, that puts Z.ai among the clearest examples of China’s AI companies turning attention into actual paying customers, and doing it at scale, not just piling up downloads, hype, and fresh funding rounds.

Bloomberg says Z.ai gets there by offering free models such as GLM-5.2, then charging for cloud access, enterprise customization, technical support, and on-premises deployments. Reportedly, 73.7% of its 2025 revenue came from deployments at state-owned enterprises and financial institutions.

On Bloomberg’s numbers, Z.ai’s open-platform ARR hit 1.7 billion yuan, about 60 times higher than the year before. API usage also kept climbing, up 400% even after an 83% price increase in Q1 2026.

There’s still a catch: Bloomberg says Z.ai posted a 2025 net loss of 4.72 billion yuan, and JPMorgan doesn’t expect the company to turn a profit until 2028. If you pay attention to AI business models, this one is worth tracking as Z.ai goes up against Moonshot AI, MiniMax, DeepSeek, Alibaba, and ByteDance, while also pushing into the Middle East, Singapore, the UK, and Malaysia.

GLM-5.2 is available as a free download. Z.ai’s paid services, meanwhile, are aimed at enterprise customers.