Chinese artificial-intelligence systems have matched the performance of Anthropic’s powerful Mythos model in some cybersecurity scenarios, security researchers told the Wall Street Journal, a development that is resetting the global technology race and pressuring the White House in its overhaul of U.S. AI policy.
Zhipu AI, also known as Z.ai, released its GLM-5.2 model this month. Security researchers said the open-weight model can match the latest U.S. models when it comes to finding security bugs, although it still lags behind Anthropic’s and OpenAI’s products on other tasks. GLM-5.2 has ranked as one of the 10 most-used AI models, according to data from OpenRouter, a company that provides access to more than 400 AI models.
In some benchmarking tests, according to the cybersecurity company Semgrep, GLM-5.2 bested Anthropic’s Claude Opus 4.8 model, which was released in May. When given further instructions, Opus 4.8 and GLM-5.2 can match Mythos in bug-finding ability, according to researchers.
On Wednesday, the Chinese cybersecurity company 360 Security Technology released a new bug-finding tool called Tulongfeng. The company said it was comparable to Mythos in finding bugs. “This kind of powerful weapon that can alter the landscape of cyberwarfare can’t remain solely in American hands,” 360 Security Chief Executive Zhou Hongyi said at a cybersecurity conference in Beijing. Zhou, an outspoken internet veteran and member of China’s top political advisory body, said China would face unacceptable risks if American entities could use advanced AI models to scan critical Chinese network systems while denying Chinese companies comparable capabilities.
“China is making sure that the gap becomes smaller and smaller over time,” said Lior Div, chief executive officer of the cybersecurity company 7AI.
Unlike models from Anthropic or OpenAI, Zhipu’s GLM-5.2 is open-weight, meaning it can be downloaded and run on any hardware and can be modified and used without supervision. The characteristic makes it attractive to users who want unfettered access but also to hackers who can run the systems in the shadows. The ability of AI systems to find bugs in software has added urgency to efforts to close vulnerabilities quickly, as some researchers have warned of a “bugmageddon.”
The U.S. government’s recent restrictions on its own advanced AI models have added complexity to the competitive landscape. On Friday, OpenAI said it was limiting access to its latest model, GPT-5.6, because of security concerns among administration officials. The company said the current case-by-case model-evaluation process was not a long-term solution.
One of Anthropic’s latest general-use models was shut down for more than two weeks after the Trump administration said no foreign entity or individual could use it because of security risks. The administration on Friday restored some access to a related Anthropic model called Mythos 5, which had previously been restricted. Among the users that had lost access before Friday’s restoration: the National Security Agency, which had been testing the tools and found them impressive in trials, according to people familiar with the matter.
Critics of the administration’s approach said restricting U.S. models while allowing the export of AI chips to China is counterproductive. “Banning Fable while selling chips China needs to develop its own version is a gift to China,” said Saif Khan, a distinguished technology fellow at the Institute for Progress think tank who worked on export restrictions in the Biden administration. The U.S. needs to maximize use of Mythos and comparable models to harden its cyber defenses while it can, he added.
“Our administration is very much focused on Chinese open-source models,” Jacob Helberg, undersecretary of state for economic affairs and a former tech adviser and investor, said in a recent interview at a summit of a coalition working to secure supply chains and counter China’s influence in AI. “It’s something that we’re tracking very closely.”
The Pentagon recently announced a deal with Reflection AI, one of the few domestic open-weight developers, for use in classified settings, along with a host of similar agreements. The deal follows other recent steps by the administration aimed at supporting U.S. open-weight developers.
AI users said that U.S. efforts to rein in the capabilities of recent cybersecurity-focused models have added to concerns that access to needed systems could eventually be cut off. “It is incentivizing companies across the globe to use cheaper but very capable Chinese open-weight models, while at the same time undermining the U.S. AI industry,” said Niels Provos, a researcher who led security teams at Google and Stripe. “I don’t understand it.”