Tech
OpenAI is scared of open-weight models. Should the US be?
The impressive capabilities of Chinese lab Moonshot’s Kimi K3, the biggest open-weight large language model, has kicked off a debate that conflates two things: the economic possibilities of American AI giants and the future of LLMs as a technology.
OpenAI’s head of strategic futures, Dean W. Ball, went so far as to argue that the US government should find a pretext to create regulatory fear, uncertainty, and distrust around the new models, since open-weight models must necessarily deter capital spending by the frontier labs.
People freaked out, with tech luminaries like Yann LeCun and Martin Casado arguing that open software can accelerate innovation and coexist with proprietary projects. Ball soon retracted his claims that a regulatory crackdown was the White House’s “best strategy” and that open-weight models necessarily slow down advances in the technology.
However, Axios reports that the Trump administration is considering banning K3 and other advanced Chinese models at the behest of American frontier labs. Another report from Politico said that the Department of Commerce would not take that step anytime soon.
The benefit for major AI companies is clear: Open-weight models, running on independent infrastructure or inside major enterprises, offers cheaper intelligence than Anthropic or OpenAI’s class-leading models. If users increasingly spend more outside the closed labs, that means smaller return on their massive investments in model training.
That view extends far beyond OpenAI. “Strong, frontier-caliber open source models will place a squeeze on the margins and will bring down the prices of the frontier companies,” Braden Hancock, the co-founder of Snorkel AI and a research partner at the Laude Institute, told TechCrunch. “It will not necessarily mean that the amount of AI usage goes down a little bit. You know, obviously, quite the opposite.”
That’s not a problem for people without shares in Anthropic and OpenAI. AI will still proliferate. So what’s the justification for the government to block Americans from purchasing something in our ostensibly free markets?
Concerns over Chinese models come in several flavors. One is protecting US data from the Chinese government; the US banned the import of modern Chinese EVs over concerns about their data gathering. But experts tend to think that open-weight models run on US servers are unlikely to leak data back to China, although it’s not impossible that such a thing could be done.
Another is that the models may have implicit bias toward the PRC — but it’s not clear what that might mean for, say, coding tasks.
A third common worry is that Chinese models lack the guardrails that the US government has mandated (through an opaque process), which aim to prevent leading US LLMs from being used to exploit closed computer systems or create weapons. However, those same guardrails may make US companies more vulnerable: David Sacks, the venture capitalist and Trump adviser, has been sharing cases of US companies turning to Chinese LLMs to close security gaps when US frontier models refuse to do the tasks.
But the most significant motivation for restricting the models is that fear that China will be able to outpace the US if the frontier labs slow down.
Sam Bresnick, a China-focused research fellow at Georgetown’s Center for Security and Emerging Technology, says the growing importance of AI to the US military operations gives the US a reason to support continued investment in AI at the frontier labs. But the whole question, he says, is fraught.
“Why should the weight of the U.S. government be aimed at protecting these these companies from competitors that are being locked out from the U.S. market based on their origins?” Bresnick asks.
Advocates for open AI say that the frontier companies are creating a false binary between innovation and closed models.
“The bigger impact of having these open source models come from China is less that they’re sneaking in back doors, and more that they are owning the innovation,” Hancock told TechCrunch. “You end up with, effectively, an expanded workforce on your model. PyTorch became the industry standard because it was open source, and so the whole community could contribute to it rather than just one company, and it grew and grew, and all the rest of the deep learning libraries kind of died in comparison.”
Hancock and other advocates fear that Chinese LLMs will become the locus of international research. Already, US graduate programs mainly build on open-weight Chinese models, and Hancock says that half of the papers students study are coming from Chinese institutions, with American frontier labs increasingly reticent about sharing their work widely.
“Restricting open models wouldn’t make AI safer,” said Clem Delangue, the CEO of Hugging Face, a platform for open AI collaboration. “It would simply hide the risks, concentrate power in the hands of a few and make it harder for the next generation of builders, researchers, academia, non-profits, governments to participate in making AI safer and more beneficial for all.”
Bresnick says that the real way to slow China would be to focus more on chip export controls. A better way to preserve US AI leadership would be to stop selling Nvidia H200 processors to China. “That,” he says, “could potentially keep us out of this thorny debate about banning open source technologies that huge numbers of US companies want to use.”
Part of the problem is that uncertainty around AI economics. “The open business model, the proprietary business model — neither one is figured out. AI companies are struggling to figure out how to make money on their tools, especially as training costs need to go up and up,” Bresnick points out.
The same challenges that play out in the US are also playing out in China, where AI companies are also struggling to generate revenue and access compute power, and the government is seen as encouraging open releases for policy reasons despite the challenge in capitalizing on them.
Some US companies, including Thinking Machines Lab and Nvidia, are trying to make a business around releasing open models. Hancock points out that Nvidia would do better “if there are dozens or hundreds of companies building AI than rather than two or three that are well capitalized enough to make their own chips,” which is one reason behind its investment in Nemotron, a collection of open models.
“The main point is the U.S. would be very well served to have its own very capable, much less expensive open models,” Bresnick said. “It just clashes with the approach the frontier labs have taken.”
With additional reporting from Rebecca Bellan.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Tech
Gritt exits stealth with $32 million for robots to build solar plants — then, everything else
One of the most important things happening on Earth today is the solar energy build-out. Around the world, companies and countries are racing to deploy solar and batteries to achieve energy independence and limit the effects of climate change.
That build-out, though, is running into a labor market challenge, with a limited supply of workers to meet a growing demand for installation. Robots could be an answer, but industrial robots have historically struggled in unstructured environments, at least until now. The latest generation of AI models may have changed that equation.
That’s the driving idea behind Gritt, a startup founded by two Carnegie Mellon-trained roboticists, CEO Puneet Puri and CTO Vishal Dugar. The company exited stealth Tuesday morning with a $26 million Series A round of funding led by Obvious Ventures, with participation from Union Square Ventures and Active Impact Investment. That brings its total funding to $32 million, following an earlier seed round backed by First Round Capital, Climactic, Congruent Ventures, and VSC Ventures. The startup is building an intelligent system to “help civilization build infrastructure faster,” in Puri’s words.
“Our thesis is that if we truly want to speed up construction,” Puri tells TechCrunch, “you need an intelligence which can work in the outdoor, chaotic environments of these construction sites, and it has to be generalizable enough that it can work in these varied environments.”
Rather than building its own robots from scratch, Gritt uses off-the-shelf hardware — thus far, rented skidders and robotic arms built by companies like Kawasaki — to build platforms that are controlled by its AI models. The first job its systems handle is unloading large, glass solar panels, carrying them toward the metal frames where they need to be installed, and positioning them on the frames with sub-millimeter accuracy so workers can fasten them.
“There are people who used to build rockets that went into space and had infinite budget for the smallest little part, and then there are people who know what it means to get into dirty, dull, and dangerous jobs and scale them like mad,” said Andrew Beebe, the partner at Obvious Ventures who led Gritt’s Series A round. “These guys are in the second camp, and that’s a special kind of entrepreneur that has the technical chops, the AI, and the machine vision skills to make it work.”
Gritt has two systems currently deployed in the field, using the data they collect to improve their behavior. Puri says that a typical eight-person crew can install 800 panels a day, but the same crew working with Gritt’s systems can install 3,000 to 4,000 panels each day.
Now, the company says it is contracted to help install 2.8 gigawatts of solar panels in the next 18 months, and that its customers include three of the top 10 U.S. power construction companies. The company hopes to be operating 48 of its systems within the next six months.
TechCrunch spoke to one Gritt customer who declined to be identified for competitive reasons, but who was enthusiastic about the system’s ability to improve his work. He expects it to be easier to work at remote sites where it is difficult to attract workers, and anticipates a reduction in injuries since workers won’t have to repeatedly lift 100-pound panels overhead.
Gritt is competing against companies with their own panel-installing robots like Luminous Robotics, Cosmic, and China’s Trinabot. Those companies are building their own hardware, rather than focusing on off-the-shelf vehicles and arms like Gritt, a difference that could shape who grows faster and with a leaner cost structure as demand grows.
Gritt wants to add new manipulation tasks to its system so it can fasten the solar panels, drill posts, and even build the racks they sit on. Longer term, it also wants to move into other common, labor-intensive construction tasks, like tying rebar before concrete is poured over it.
What’s enabled the startup to pursue this vision? Mainly, the rise of new AI models, the founders say.
“Making a system for one solution was still possible to some extent five years ago, right?” Puri said, but AI is now making that work generalizable — the same underlying pipeline can be reused and improve across tasks. As an example, he noted that training the system to stack cinder blocks took weeks, while a similar demo with rebar tying took just a day using the same software.
But training new tasks is just the beginning of Gritt’s vision. The founders believe the suite of sensors and intelligence its systems bring to worksites can do more than install panels; it can boost management and decision-making. For instance, they imagine their system noticing a trench is open while a storm approaches, allowing it to alert workers to cover it before rain damages components, or flagging missing inventory.
“Gritt becomes now this layer of physical AI, which is doing this dextrous, labor-intensive task, plus it can help you take decisions on the site,” Puri said.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Tech
Bluecore Energy raises $10M to build portable nuclear reactors on barges
Maritime nuclear energy startup Bluecore Energy on Tuesday said it had raised $10 million in a pre-seed funding round that was led by Slauson & Co.
Founded seven months ago by Kofi Asante, who previously worked with Uber Freight, Bluecore is building small nuclear reactors (SMRs) on floating barges with an aim to provide clean power to ports and nearby infrastructure. The reactors heat water and transfer the resulting steam into a generator, which then spins a turbine to generate electricity, Asante explained. The system is water-cooled in a closed-loop.
The energy expected to be produced on Bluecore’s barges can be moved by ship to its next location, reducing the emission involved in its transport to zero, Asante claims. Plus, he said the entire system behind the nuclear power plant only needs to be refueled once every few years.
Bluecore’s barges can also be docked near communities, and can connect to the power grid via subsea cables. The goal is to try to power the “equivalent of approximately 15,000 homes or scale to meet the power needs of a major port,” he told TechCrunch.
“We are able to utilize existing water-cooled nuclear technology that has been operating for over 70 years,” he said. “With a production line of small modular reactors that can be rapidly deployed on water, there is a pathway to provide clean energy to the majority of the country.”
Bluecore will be using the fresh capital to deploy its product. It has already secured a port terminal, barge, and test reactor pressure vessel, Asante said. “The test vessel allows us to simulate flow with water, which is the cooling source of the system. We are combining hardware with software testing to validate and verify the foundation of our design,” he added.
The startup is working with regulatory agencies to “embed the safest design decision” into its first product. Asante said the startup is building many layers of “safety and redundancy,” like having the uranium clad and protected in a thick steel pressure vessel and then padded with concrete shielding and steel lining.
Asante is hoping Bluecore may be able to help with the increasing power demand sparked by the ongoing data center buildout. “AI data center execs have shared with me that they would not need to pull water or energy from communities around them if they are able to receive their own source of electricity and have access to water that is provided at sea,” he said.
Other investors in the round include Harlem Capital, Precursor Ventures, Ripple co-founder Chris Larsen, and actor Kevin Hart’s HartBeat Ventures, as well as a few angel investors.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Tech
Music streamer Deezer says more than 50% of daily uploads are AI-generated
Music streaming company Deezer has been tracking the number of AI-generated tracks uploaded on the platform since last year, and the number has constantly gone up. Today, the company said that AI music now represents more than 50% of downloads.
Deezer said that AI-generated track uploads were at a peak in June 2026, representing a monthly average of 90,000 tracks per day.
The rapid rise of AI-generated music has forced streaming services to decide how much of it they want on their own platforms. There is no single consensus yet on that front. Some take strict steps, like Bandcamp banning such tracks or Tidal cutting off monetization. Meanwhile, Apple Music has a voluntary AI-tagging system, and Spotify developed its own policy about how much AI was used in music-making.
Deezer’s latest move on this front will involve taking down AI-generated tracks that haven’t been streamed in the past six months or are involved in fraudulent streams to drive up revenue.
“Deezer has been at the frontline of fighting fraud and reducing payment dilution related to AI music for almost two years. Now that half of all daily uploads are AI-generated tracks, we are taking additional steps to safeguard the rights of artists and songwriters, while maintaining focus on music that fans actually love,” Deezer CEO Alexis Lanternier said in a statement.
The streamer first released stats around AI music uploads in January 2025, when the daily upload volume was around 10,000 tracks, or 10% of daily uploads. The number grew to 20,000 tracks, or 18% of daily uploads, in April 2025. It then climbed to 30,000 tracks, representing 28% of daily uploads in September 2025, followed by 50,000 daily uploads, or 34% of daily uploads, in November 2025.
This year, it grew again to 60,000 tracks, or 39% of daily uploads, in January 2026. As of April 2026, the figure reached 75,000 tracks, or 44% of daily uploads.
Deezer started labeling AI music on its platform last year, and said that its detection tech can also identify tracks generated with models from Suno and Udio, AI-music startups that are embroiled in copyright lawsuits. Earlier this year, Deezer made its detection tech available to other platforms, but it’s not clear if any of the major platforms are using the tool just yet. Last month, it also released a tool that can sift through Apple Music and Spotify playlists for AI-generated tracks.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Guardians go deep seven times as they club Twins
Rays end losing streak, shut down struggling Jays
