China’s Z.ai GLM-5.2 Narrows Gap With US Models on Cybersecurity

Zhipu AI's open-weight GLM-5.2 matches Anthropic's Mythos on some bug-finding tasks, raising U.S. national security concerns.

China’s Zhipu AI, operating under the brand Z.ai, has released GLM-5.2, an open-weight large language model that some researchers say matches Anthropic’s Mythos in certain bug-finding and cybersecurity scenarios, according to reporting by The Verge. [1]

The model still lags behind offerings from Anthropic and OpenAI on more general tasks, but the narrowing of the gap in cybersecurity-specific capabilities has drawn attention from U.S. officials. [1]

Because GLM-5.2 is an open-weight model — meaning its underlying parameters are publicly available and can be downloaded and run locally — it can be operated by anyone on readily available hardware. [1] That openness gives developers and power users deep access and flexibility, but it also means the model can be run with little oversight, raising concerns about potential misuse by bad actors. [1]

The Trump administration has been working to restrict China’s access to advanced U.S. AI models, including Anthropic’s Mythos and Fable, as well as the specialized hardware required to train and run them. [1] The administration views models capable of identifying software vulnerabilities as serious national security threats. [1]

Separately, OpenAI recently unveiled GPT-5.6, which has also prompted concerns about misuse potential, leading to limited access to that model. [1]


Sources

  1. The Verge — China’s Z.ai claims it can match Mythos on cybersecurity

This article was drafted with AI from the cited sources and checked against them before publication. Spot an error? Let us know.

The Trump administration forced Anthropic to pull two new models citing national security, but researchers and the company itself are pushing back on the rationale.

The United States government has ordered Anthropic to withdraw its two newest artificial intelligence models, Fable 5 and Mythos 5, from availability, citing national security concerns. [1]

The action followed reports that Amazon researchers allegedly discovered a method to bypass the safety guardrails built into Fable 5. [1]

Anthropic has responded by noting that the same jailbreak techniques — methods used to circumvent a model’s built-in restrictions — exist in other AI models as well, not just its own. [1]

Cybersecurity researchers have also pushed back, signing an open letter characterizing the government’s move as dangerous. [1]

The ban has immediate practical consequences for developers building products and services on Anthropic’s platform, as access to the two newly released models has been suspended. [1]

The episode is being described as the latest development in what observers have characterized as a fraught relationship between Anthropic and the Trump administration. [1]

Amazon’s chief executive reportedly raised concerns about the Anthropic models before the government crackdown took effect. [1]

Some analysts have suggested the ban could paradoxically benefit Anthropic’s public profile, a dynamic that has drawn attention ahead of the company’s anticipated initial public offering (IPO). [1]


Sources

  1. TechCrunch — Is the US government’s Anthropic ban accidentally helping the brand?

This article was drafted with AI from the cited sources and checked against them before publication. Spot an error? Let us know.

Alleged China connections at SK Telecom and security flaws in Fable 5 led the White House to pull Anthropic's Claude Mythos and Fable 5 models offline.

A combination of national security concerns and reported safety vulnerabilities prompted the White House to force Anthropic to take two of its AI models — Claude Mythos and Fable 5 — completely offline, according to a report by WIRED cited by The Decoder. [1]

The crisis began when South Korean telecom company SK Telecom was found to have access to Mythos through Anthropic’s partner program, known as Project Glasswing. [1] US officials grew alarmed over what they characterized as SK Telecom’s alleged ties to China, and the White House directed Anthropic to cut off the company’s access. [1] Anthropic complied immediately. [1]

SK Telecom denied having any China connections when contacted by a Korean newspaper. [1] However, the company is part of the broader SK Group conglomerate, which has significant business interests in China and held a stake in state-owned Chinese carrier China Unicom until 2009. [1]

The situation escalated further when Amazon and other companies flagged security flaws in Fable 5 that could be exploited to bypass the model’s safety restrictions. [1] The two incidents together caused the White House to lose confidence in Anthropic and order both models taken fully offline. [1]


Sources

  1. The Decoder — Alleged China ties at SK Telecom alarmed US officials and triggered Anthropic crisis

This article was drafted with AI from the cited sources and checked against them before publication. Spot an error? Let us know.

The Justice Department argues that shutting down xAI's Colossus 2 facility would harm military operations that rely on the Grok AI model.

The U.S. Department of Justice (DOJ) has filed a motion to dismiss an NAACP lawsuit targeting xAI’s unpermitted gas turbines, arguing that the case poses a threat to national security by potentially cutting power to AI infrastructure used by the U.S. military. [1]

In its filing, the DOJ contends that the suit “threatens American national, economic, and energy security by seeking to shut off the power supply for artificial-intelligence innovation that supports the Department of War’s military operations.” [1]

The national security argument rests on a declaration from Cameron Stanley, Chief Digital and Artificial Intelligence Officer at the Department of Defense, who states that Grok — xAI’s large language model and chatbot — is one of only four AI models that “support mission-critical operations across Secret and Top-Secret classified networks,” including recent strikes against Iran. [1]

The NAACP brought the lawsuit over xAI’s operation of unpermitted gas turbines at its Colossus 2 data center in Southaven, Mississippi. [1] According to the Southern Environmental Law Center, the number of turbines at the site has grown from 27 to 57 since April, resulting in a 111 percent increase in nitrogen oxide emissions. [1]

xAI, which is part of SpaceX and owned by Elon Musk, operates Colossus 2 as one of just two major data centers in its portfolio. [1]


Sources

  1. The Decoder — DOJ invokes national security to defend xAI's unpermitted gas turbines in NAACP lawsuit

This article was drafted with AI from the cited sources and checked against them before publication. Spot an error? Let us know.

Export restrictions on Anthropic's Mythos model were driven partly by fears a China-linked group had accessed the system, according to a Semafor report.

The White House reportedly suspects that a group linked to China gained access to Anthropic’s powerful Mythos artificial intelligence model, a concern that partly drove the government’s decision to impose export restrictions on the system, according to a report from Semafor cited by The Verge. [1]

If confirmed, the breach would carry significant national security implications. [1] One specific concern is that the Chinese government could attempt to replicate the model’s capabilities through a technique called distillation — a process in which a less advanced “student” AI is trained on outputs from a more capable model in order to reproduce its behavior. [1]

The White House has not confirmed the Semafor report. [1] A post on X by Trump advisor David Sacks did not mention China; instead, Sacks focused on reported vulnerabilities that would allow Fable and Mythos to be jailbroken — a claim Anthropic has denied. [1] An Anthropic spokesperson told Semafor that the government did not raise China during its discussions about export controls, and Anthropic did not reply to a separate request for comment from The Verge. [1]

The suspected China-linked access would not be the first known security incident involving Mythos. [1] Anthropic has previously stated that Mythos is too dangerous and powerful for public release, yet a Discord group reportedly had access to the model for approximately two weeks before Anthropic discovered the breach and revoked it. [1]


Sources

  1. The Verge — China may have accessed Mythos

This article was drafted with AI from the cited sources and checked against them before publication. Spot an error? Let us know.

A government export control directive citing national security forced Anthropic to block all global access to its two newest frontier models, affecting paying enterprise customers and its own employees.

The U.S. government on Friday evening issued an export control directive ordering Anthropic to immediately suspend access to its Claude Fable 5 and Claude Mythos 5 models for all foreign nationals, citing national security concerns. [1,2] In response, Anthropic blocked all public access to both models globally — including for paying enterprise customers and Anthropic’s own employees. [2]

The shutdown came just three days after Fable 5 and Mythos 5 had been publicly released. [2] Users with active sessions on either model are now receiving errors, and new queries are being automatically routed to older models such as Opus 4.8. [2] Anthropic said in a blog post that access to all other Anthropic models remains unaffected. [2]

Anthropic said it is complying with the order but pushed back on its basis, stating that the government “did not provide specific details of its national security concern.” [1] The company said any evidence of a potential jailbreak was conveyed verbally rather than in writing. [1] In its public statement, Anthropic described the government’s disclosure as limited: “To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws.” [2]

Anthropic further argued that the vulnerabilities identified were minor and not unique to its models, explicitly naming rival OpenAI’s GPT-5.5 as a system where similar capabilities are available. [1,2] In its statement, the company said: “We have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result.” [1] Anthropic also warned that pulling a commercial model over a non-universal jailbreak sets a regulatory standard that could “essentially halt all new model deployments for all frontier model providers.” [2]

The government action follows a viral jailbreak of Fable 5 published on X on June 10 by a user known as “Pliny the Liberator,” who claimed to have bypassed the model’s safety guardrails to extract instructions for cyber exploits, explosives, and chemical synthesis pathways. [2] The attacker reportedly used a multi-agent technique combining Unicode characters, homoglyphs, Cyrillic script, long-context reference tracking, and a method of breaking harmful requests into innocuous fragments before reassembling them using a previously jailbroken Opus model. [2] Anthropic did not confirm whether this specific jailbreak triggered the government order. [2]

Anthropic said it believes the shutdown is “a misunderstanding” and is working to restore access as soon as possible, and apologized to its customers. [2] The company also outlined steps it had already taken to safeguard the models, including working with U.S. and UK governments and updating its data retention policy to help track attempts at malicious use. [1]

The incident is not the first time Anthropic has faced U.S. government friction. In March 2026, Secretary of Defense Pete Hegseth labeled Anthropic a “supply chain risk” after the company declined to allow the military to use Claude for mass domestic surveillance and lethal autonomous weapons without safety restrictions, resulting in a prohibition on Anthropic’s use across defense supply chains. [2]

The sudden blackout has prompted discussion in the enterprise and developer community about the risks of relying on a single cloud-based AI provider. AI founder Alex Finn posted on X calling the Anthropic shutdown a “wakeup call” and urged developers to run local models on their own hardware, writing: “No company or government will EVER be able to take away your local models.” [2] Chinese open-source AI provider MiniMax highlighted the availability of its open-weights M3 model as a contrast to the centralized vulnerability exposed by the Claude shutdown. [2]

The episode illustrates a practical tension for enterprise buyers: closed-API frontier models offer the most advanced reasoning and agentic capabilities but depend entirely on vendor and regulatory continuity, while locally hosted open-weights models provide operational independence at the cost of cutting-edge performance. [2]


Sources

  1. The Verge — Anthropic cuts off Fable 5 and Mythos 5 access following government order
  2. VentureBeat — Anthropic blocks all public access to Claude Fable 5, Mythos 5 following US government order — what enterprises should do

This article was drafted with AI from the cited sources and checked against them before publication. Spot an error? Let us know.