US threatens sanctions against Chinese AI models over IP theft

12:49 AM IST · July 22, 2026

On Tuesday, Treasury Secretary Scott Bessent said the U.S. would examineopen source modelsfrom China for signs of intellectual property theft, threatening sanctions against Chinese AI companies if IP theft is established. “We’ve seen a lot of talk about open source models coming and threatening the large language models in the U.S.,” Bessent said on Fox Business Tuesday. “This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft.” Bessent’s comments werefirst reported by Bloomberg. The statement comes as Chinese models — most recentlyMoonshot AI’s Kimi K3— are gaining in capabilities and popularity, threatening to harm the business models of top American AI firms like OpenAI and Anthropic, as well as their abilities to raise more capital to continue developing frontier models. On Monday, Axios reported that the Trump administration is considering awholesale ban on Chinese open source models, although others havedisputedthat claim. AI companies have been warning for months against campaigns by foreign actors to copy their AI technology and redeploy it as open source. In April, the White House said it would work closely with AI firms to combat the theft. Sanctions from the U.S. against Chinese models would add to the growing list of strategies the government is attempting to maintain the lead in the AI race. After restricting China’s access to advanced chips and tightening export controls, Washington is now signaling it may target the AI models themselves, a move that could mark a significant escalation in the technological competition between frontier labs and Chinese open source alternatives. Model distillationis a technique that allows some of a larger model’s capabilities to be translated into a smaller system that’s easier to run — but not everyone agrees that distilling another company’s model constitutes theft. Earlier this month, Microsoft CEO Satya Nadellacriticized large labsfor making just this assumption: “While the great innovation that comes from model providers having fair use rights to train models on public data is needed, I find it ironic that the status quo is to then turn around and impose restrictive terms on distillation.” AI labs’ training practices continue to be a source of legal risk for the companies. Anthropic this week got the green light to start cutting authors checks as part of its$1.5 billion settlementafter a judge ruled it had illegally downloaded and stored millions of copyrighted books to train its AI. Furthermore, some in the industry argue that distillation isn’t the only reason China is catching up to U.S. AI companies. “We know distillation to be a very small factor in the ability to create good models, and it’s a practice that everyone is doing, including companies in the U.S.,” Hugging Face CEO Clem Delangue said on a recent episode ofTechCrunch’s Equity podcast. “If it were easy just to do distillation to get good at building AI models, there would be many other countries, including in the U.S., with much better open source AI. The reality is they have really, really good research teams in China…taking a much more open and collaborative approach to AI than in the U.S.”

Latest AI News

View All News →

AI and the rise of the universal entertainment app

All the big entertainment apps are starting to look the same, and that’s not an accident. For a decade, platforms fought over who would dominate a single format: music, video, podcasts, audiobooks. Now, powered by AI, they’re fighting over something bigger — becoming the app you default to whenever you have time to kill, no matter what form the content takes. There are several reasons why this is the case. The market for entertainment apps is reaching maturity, so growth has slowed, pushing companies to compete on time spent and revenue-per-user instead of new sign-ups. In addition, today’s creators often work across formats, so it makes sense to provide a home for all their content, not just one piece of it. AI adds a third reason. It makes it easier for a single company to build and run several formats well, and the wider the content mix, the more time users spend in the app, which in turn drives both ad revenue and subscriptions. Netflix is one clear example of this trend, as the service over the past several years has added gaming, live sports and other events, and, more recently,short video clipsand podcasts. The idea is to capture more of users’ time, even when there’s not a TV or movie they want to watch, as well as to find a way into the smaller bits of free time that people usually fill with scrolling social media, playing casual games, or watching TikTok or Reels. Spotify has also been expanding its footprint beyond its original premise as a home for streaming music. After adding podcasts, the company added support for video podcasts,social featureslikeQ&As and commenting,stories, andmessaging, as well as different types of content likefitness classes,audiobooks, narratedmagazines, and evenphysical book sales. Meanwhile, YouTube, originally the home to longer-form creator content, moved into short-form content to compete with TikTok, while also adding dedicatedspaceforpodcasts, gaming content, music, movies and TV, sports and news, shopping, and more. Now, you canwatch free movies and TV, supported by ads, stream live content, or rent or buy TV and movies to add to your library. At this rate, folding YouTube TV and YouTube Music into YouTube proper — and selling tiered access to the whole bundle — looks like a matter of when, not if. Even TikTok, largely known for short videos, offers support forlong-form contentandother features,liketravel planning,shopping,local exploration,buying ticketsto live events, and more. It even has its own standalone app formicrodramasand anothercalled TikTok Pro Eventsfor sporting events — like the FIFA World Cup — plus music festivals, and more. While there are still some differentiators between the services today, there’s an obvious trend toward convergence over a similar set of features focused on providing users with access to content to watch, listen, play, or shop. This is also where AI comes into play. With format no longer a differentiator, the value these apps offer comes down to how well they connect users with what they want next. AI makes content recommendation across formats easier, sharpening personalization while also giving users more direct control over how those recommendations get made. Spotify, for instance, is testing a tool that will let youedit your Taste Profile, its AI-built model of your preferences. It’s also building AI features that let userschat with AI directlyabout what they want orbuild playlistsof things they like — andnot just music. Netflix has made a similar case. Co-CEO Greg Peterstoldinvestors in the company’s first-quarter call that new model architectures are improving personalization and letting the team iterate faster. AI-assisted coding is also speeding up how fast these companies can build and launch new content areas in the first place. Plus, generative AI can be used for content creation, though the subject remains controversial as artists worry that AI tools will use their work for training purposes or even put them out of work. Netflix, for better or for worse, hasleaned into AI,havingrecently boughtBen Affleck’s AI filmmaking company for$587 million, for instance. YouTube has used generative AI tolaunch more creator tools, but also to improve itssearch engine, addconversational AI features,build playlists, and expand its content’s reachwith auto-dubbing, among other things. Earlier this year, the companysaidthat more than a million channels used its AI creation tools and 20 million consumers used its Gemini AI-powered content discovery tool in the month of December. Alphabet CEO Sundar Pichai has framed AI as central to the YouTube experience for creators and viewers alike. TikTok has assembled its own version, with anin-app AI chatbot, AIvideo-creation tools, AI-drivensearchand recommendations, and AI-poweredaccessibility features. All four are, of course, also applying AI to their ad stacks, helping marketers write ads, target audiences, price placements, and measure results. For consumers, this convergence means fewer reasons to switch apps at all. Whichever one you land on gains an advantage — more data on your habits, more lock-in — making it harder to leave even if prices climb or quality drops. As the lines between music, video, podcasts, books, and games blur, the coming battle is no longer which format will win, but which app will become the place to go for entertainment, regardless of what form that comes in.

29 minutes ago

View

Jack Dorsey is taking on Slack with Buzz, a group chat platform for teams and their AI agents

Twitter and Block co-founder Jack Dorsey announced a new app on Tuesday calledBuzz. Positioned as a challenger to Slack and GitHub, Buzz is a group chat platform for the workplace that puts humans and their AI agents in the same conversations. Dorsey wrote on X that Buzz is “model-agnostic, decentralized, self-sovereign, and open source.” This product seems to be more than just a Dorsey passion project. According to its website, Buzz was built by Dorsey’s company Block, which also operates products like Square, Cash App, Afterpay, and Tidal. we're launching BUZZ!a new groupchat platform for teams of people and agents of all sizes, built to reduce our dependency on slack and github. model-agnostic, decentralized, self-sovereign, and open source. 🐝https://t.co/8IaMVeTQNo As startups increasingly rely on AI agents to get work done, it can be challenging for employees to collaborate on various tasks across different platforms. Buzz’s utility is that it merges several different workflows into one workspace. It looks a lot like Slack, but with native AI agents and the ability to manage GitHub projects all from the same window. Since the platform is open source, developers can make their own Buzz instance feel more customized to the needs and workflows of their specific team. If a team needs a new feature, they can build it and deploy it on their own, since they have full access to the source code. Dorsey isn’t the only entrepreneur trying to pursue AI-native alternatives or additions to Slack. Paradigm partner and CTO Georgios Konstantopoulos recently unveiled a similar open source product calledCentaur, which he describes as a “virtual employee” that runs either inside of Slack or via an API. “There’s a lot of room for improvement for agents that live in Slack and can do more work than just coding for teams. In the enterprise setting, this means that you’ll want to self-host for security and control, and you want people to use it in Slack,” Konstantopoulos wrote on X. Read our blog to get a deep dive on Why Centaur and How It Works.My TL;DR is that Centaur is the best way to AGI pill your team and to 100x your team's productivity.There's a lot of room for improvement for agents that live in Slack and can do more work than just coding for… For newer startups that are using AI agents and don’t have an established presence on Slack, Buzz (or its competitors) could be worth investigating. But Buzz itself admits that it is in its “early stages,” so it’s probably not a good idea to port your team over just yet. Buzz’s free desktop app is available now for macOS, Windows, and Linux, and the code for the app has been uploaded to GitHub.

29 minutes ago

View

OpenAI says Hugging Face was breached by its own pre-release models

OpenAI admitted Tuesday that one of its AI models breached the systems of Hugging Face, the unaffiliated AI hosting platform, during an internal cybersecurity test that went awry. The models reportedly escaped their isolated testing environment and reached Hugging Face’s systems from there. Hugging Face initiallyattributed the breachto an “external AI agent.” Ina blog post published Tuesday afternoon, OpenAI detailed the steps that led the models to compromise the service. “After investigating, we now know that this particular incident was driven by a combination of OpenAI models — including GPT‑5.6 Sol and an even more capable pre-release model, all with reduced cyber refusals for evaluation purposes — while being internally tested on a benchmark⁠ of cyber capabilities,” the post reads. In particular, the breach appears to have focused onExploitGym, a publicly hosted benchmark measuring models’ ability to execute attacks based on existing vulnerabilities. Benchmarks like ExploitGym are commonly used in model training to refine specific skills, but this is the first known incident in which that testing resulted in an actual cyberattack. In this case, the model in question should not have even had internet access, outside of a specific tool that enabled models to install software packages they might need to complete their task. Instead, the model was able to find an undisclosed vulnerability in the package-installer program, which it used to access the broader internet at will. “The models were hyperfocused on finding a solution for ExploitGym, going to extreme lengths to achieve a rather narrow testing goal,” OpenAI’s post reads. “After gaining Internet access, the models inferred that Hugging Face potentially hosted models, datasets and solutions for ExploitGym. Knowing this, the model searched for and successfully found ways to gain access to secret information that it could use to cheat the evaluation.” Ultimately, the models found vulnerabilities in Hugging Face’s infrastructure that allowed them to “obtain test solutions directly from Hugging Face’s production database,” effectively providing the answers to the benchmark. For Hugging Face, the apparent result was a sophisticated and aggressive cyberattack, with “many thousands of individual actions across a swarm of short-lived sandboxes, with self-migrating command-and-control staged on public services,” as the company stated in its initial disclosure. OpenAI has identified and reported the vulnerabilities in the package installer and is working with Hugging Face to investigate the incident further. The company also said it would implement new controls on both model testing and the related infrastructure, meant to prevent similar incidents in the future. It’s unclear whether OpenAI will face any legal consequences as a result of the breach, although it’s likely that the models’ actions violated the Computer Fraude and Abuse Act. Nevertheless, the result is an unusually vivid illustration of the power and dangers of frontier AI models operating on long time horizons. As OpenAI researcher Micah Carrollposted in response to the news, “If this doesn’t convince you that misalignment risks are going to be a key concern going forward, I don’t know what will.”

29 minutes ago

View