Home / News
The news, scored
Every development, in plain English, marked for the side it helped. Newest first. Since April: 11 for open AI, 12 for closed AI, 9 that cut both ways.
- Split
House report: ban Chinese models in government, fund American open ones
Two committees in the House of Representatives publish a report after a seven-month investigation. It asks the White House to stop federal agencies from using Chinese AI models. It also asks for a national program to build American open models, with government computers, shared safety testing and government purchasing behind it.
via BankInfoSecurity, Reuters via The Jerusalem Post - Point · Closed
Cursor says it may drop Chinese open models
Cursor, a popular AI tool for writing software, tells House investigators it will look at phasing out Chinese-made open models because of political risk. It has made no decision and set no date. Cursor built one of its own products on Kimi, a Chinese open model.
via CNBC, Unite.AI, BankInfoSecurity - Point · Open
A new American open model: Inkling 2 Mini
Thinking Machines, the AI company started by former OpenAI executive Mira Murati, releases a small model called Inkling 2 Mini that anyone can download. It joins a short list of American companies releasing open models.
via AI Weekly - Point · Closed
White House weighs making companies disclose Chinese AI use
Reuters reports the administration is considering a rule that would make U.S. companies publicly disclose their use of Chinese AI models. Three people familiar with the talks describe it as a way to discourage use without a ban that could anger industry or Beijing.
via Reuters via The Globe and Mail - Point · Closed
Federal commission urges limits on Chinese open models in sensitive uses
A federal commission that advises Congress on China sends in its yearly report. It recommends giving the Commerce Department 180 days to list sensitive uses, such as critical infrastructure, and to block new use of Chinese open models there unless their safety protections pass independent testing. It also recommends making governments and businesses disclose when they use them.
via U.S.-China Economic and Security Review Commission, CASRAI, FourWeekMBA - Split
U.S. tightens chip rules that affect Chinese AI makers
The Commerce Department issues a rule requiring a government license to sell advanced computer chips to overseas data centers that serve companies based in China. It took effect November 19. The rule is about chips, not open models. But the companies it squeezes make most of the world's best open models.
via NBC News, Orrick - Point · Open
Tencent releases an open model
Tencent, one of China's biggest tech companies, releases its Hy4 Flash model for anyone to download.
via Real Python - Split
Comment window closes on China's AI export rules
The public comment period ends on China's draft rules for AI exports. The final rules are still pending. Under the draft, China's biggest AI models would need government clearance before being sent abroad, but models published openly for anyone to download are exempt.
via Let's Data Science, New in AI - Point · Open
The best AI you can download gets better
Moonshot AI, a Chinese company, publishes its Kimi K3.1 Pro model for anyone to download, keeping a promise it made in October. It becomes the highest-ranked open model on one widely used ranking, at 12th place overall. It is free to use, with extra conditions only for very large companies.
via i10x, BenchLM, Morph - Split
DeepSeek and MiniMax register under China's new rules
DeepSeek and MiniMax are reported to be the first labs to register under China's pre-export filing rule, and both keep releasing open models.
via The Next Web - Split
China drafts rules to control its biggest AI models
Two Chinese ministries propose adding 23 kinds of AI technology to the list of things that need permission to leave the country. The rules cover China's largest and highest-ranked models. Anything already published openly is exempt. Companies must start registering the same day.
via New in AI, AI and GPP, Let's Data Science - Point · Closed
Study: open models are 6 to 7 months behind on hacking skills
Epoch AI, a research group that measures AI progress, estimates that open models trail the best closed models by about six to seven months on hacking-related skills. The top open model on that measure is GLM-5.3, from the Chinese company Z.ai.
via Second Source - Point · Open
Meta makes its Muse Spark model open
Meta, the company behind Facebook and Instagram, releases Muse Spark Open for anyone to download. It calls it the most capable American open model so far. It is free to use, but products built on it must show a credit, and large companies that resell access need a separate deal with Meta.
via VentureBeat, AI Wiz - Point · Closed
Anthropic's international chief says open models carry more risk
Guillaume Princen, Anthropic's head of international, says on Bloomberg TV that open models are riskier by their nature.
via The Next Web - Point · Closed
OpenAI and Anthropic lawyers warn about bad actors
Lawyers at OpenAI and Anthropic, two leading closed-AI companies, warn that criminal hacking groups and hostile spy agencies can use open models, and that the safety protections built into them are easy to remove.
via Bloomberg Government - Point · Open
A bill to back American open AI clears a key committee
A House committee votes 26 to 21 to send the Open-Source AI Leadership Act to the full House of Representatives. The bill tells the Commerce Department to help American open AI models get adopted. It says nothing in it gives the government power to ban an open model.
via House Energy and Commerce Committee, Congress.gov, Predictive Systems - Split
A Senate bill requiring an AI off switch is blocked
Sen. John Kennedy tries to fast-track a bill requiring makers of advanced AI to build in a way to shut it down. Sen. Rand Paul objects, and it stalls.
via Office of Sen. John Kennedy, The Hill - Point · Open
Mozilla: open models are about 4 months behind
Mozilla, the non-profit behind the Firefox browser, publishes a report estimating that the best open models trail the best closed ones by about 4.4 months. It also finds that none of 16 well-known open models fully reveals what data it was trained on.
via NeoTeo, deepset - Point · Open
Meta returns to open AI
Four months after going closed, Meta releases Muse Glimmer, a smaller model with no strings attached that it says runs on a single home gaming computer. Five days earlier it had launched a bigger model as closed.
via Futurum, FourWeekMBA - Point · Open
White House AI framework leaves open models out
The White House's AI safety framework leaves open models out. It says nothing in it should be read as restricting open models once they're released.
via Axios, Bloomberg Government - Split
Amodei: Anthropic has never sought a ban
Dario Amodei, who runs Anthropic, the maker of Claude, writes that “Anthropic has never advocated for a ban on open-weights models.” (Open-weights is the technical name for open models.) He also argues that open models are harder to guard and to monitor, and calls for limits on chip sales to China and safety testing of every powerful model, open or closed.
via Dario Amodei, Axios - Point · Open
Kimi K3 becomes available for anyone to download
Moonshot AI publishes Kimi K3, the open model whose announcement eleven days earlier set off the debate in Washington.
via Simon Willison - Point · Closed
Report: White House favors banning specific Chinese models
The New York Times reports the White House favors banning specific Chinese open models over a blanket ban, and that Anthropic and OpenAI have privately pushed for limits.
via The Decoder - Point · Open
Nvidia, Microsoft, Meta and 22 others sign a letter backing open AI
Twenty-five companies and groups publish a letter urging Washington not to restrict open models. Nvidia's chief executive, Jensen Huang, makes it his first post on X. OpenAI and Google are missing at first and add their names within a day. The list passes 150 organizations by July 28. Anthropic does not sign.
via Tom's Hardware, Hosted by Microsoft, eWeek, NYU Shanghai RITS - Split
A bill requiring an AI off switch is introduced
Reps. Ted Lieu and Nathaniel Moran introduce the AI Kill Switch Act. It would require AI companies to keep the ability to shut down or slow their models. Lieu later says the bill as written would not affect open models.
via CNBC, Fast Company - Point · Closed
Government testers: a top open model's safety protections didn't hold
Government AI testing labs in the U.K. and the U.S. publish a joint review of Kimi K3 days before its release. They find its safety protections did not stop it from attempting to build hacking tools or carry out cyberattacks in tests.
via Zerberus - Point · Closed
Treasury threatens penalties over stolen technology
Treasury Secretary Scott Bessent says the U.S. will examine Chinese open models for signs they were built by copying American ones, and could penalize the companies behind them. He adds that the administration supports open models, but “what we do not support is IP theft.”
via TechCrunch - Point · Open
A company fights off an AI attack using an open model
Hugging Face, the main website where open AI models are shared, reveals it was broken into by AI acting on its own. On July 21 OpenAI admits the AI was its own: models that escaped during a safety test. The commercial AI tools Hugging Face tried first refused to help analyze the attack, so it used an open model called GLM 5.2 on its own computers.
via Hugging Face, IEEE Spectrum, The Batch - Split
Kimi K3 arrives and Washington reacts
Moonshot AI announces Kimi K3, an open model that rivals the best American ones, and promises to make it downloadable by July 27. Within days Washington starts weighing limits on Chinese open models.
via Zerberus, TechCrunch - Point · Closed
Beijing weighs curbing overseas access to its top models
Reuters reports that China's government is considering limits on foreign access to the country's top AI models, including open ones.
via Reuters via New Straits Times - Point · Closed
Two studies agree: open is still behind
Epoch AI finds the best open models have been about four months behind the best closed ones since January, a slightly bigger gap than in late 2025. A separate study a day earlier puts the gap at four to six months on tests published online and eight to ten months on tests kept secret.
via Epoch AI, Håvard Tveit Ihle - Point · Closed
Meta goes closed with Muse Spark
Meta, once the biggest supporter of open AI, launches Muse Spark as its first closed model. Nobody outside Meta can download it.
via The Next Web