Tag: Z.ai

  • China Considers New Restrictions on Overseas Access to Advanced AI Models

    China Considers New Restrictions on Overseas Access to Advanced AI Models

    Beijing Reviews Frontier AI as Strategic Technology

    China is evaluating potential new controls that would restrict overseas access to its most advanced artificial intelligence models, according to sources familiar with recent government-led discussions. The talks signal that Beijing now views frontier AI systems as strategic assets rather than mere commercial software, raising the stakes for global developers and businesses that rely on Chinese AI technology.

    Over the past month, officials from China’s Ministry of Commerce and the state planning agency met with major technology firms including Alibaba, ByteDance, and Z.ai to discuss possible limits on the use of top AI models outside the country. The proposals cover both closed-source systems and open-weight models, which allow users to download and modify parts of the technology. Sources told Reuters that the plans remain under review and that no final decision has been made.

    National Security Concerns Drive the Initiative

    The Ministry of Commerce led the meetings, with one proposal suggesting that leaks or theft of proprietary AI technology could be treated as an offense under China’s national security law. Officials also discussed tightening controls on who can fund domestic AI startups, adding another layer of oversight around sensitive technology and data.

    Chinese authorities are reportedly worried about advanced US AI systems, such as Anthropic’s Mythos model, and their potential use in finding software vulnerabilities. This mirrors US concerns that foreign actors could misuse frontier AI tools for military or intelligence purposes. The talks reflect a growing mutual anxiety between the two countries over AI security.

    Scope of Restrictions Remains Unclear

    Alibaba’s Qwen, ByteDance’s Doubao, and Z.ai’s GLM-5.2 are among the Chinese models that have gained global attention for their lower costs and strong performance. However, officials indicated that any new rules may apply only to future models, leaving existing systems unaffected. The scope and timeline remain uncertain, creating ambiguity for companies that integrate Chinese AI tools into software, research, automation, and customer service.

    Legal experts in China have already discussed a tiered approach for open-source AI. Under one proposal, basic tools would require simple filing, more powerful systems would face security reviews, and the most sensitive frontier models could be limited to domestic use only.

    Global Developers Face Potential Disruption

    Chinese AI models, particularly DeepSeek’s R1, have become a low-cost alternative for developers worldwide. If Beijing proceeds with restrictions, foreign companies may lose access to affordable AI options, potentially shifting to US or European alternatives or seeking domestic solutions. Costs could rise for businesses that have come to rely on Chinese AI systems.

    For now, formal rules have not been announced. Sources said it is not clear when, or even whether, the curbs will take effect. The global AI community continues to watch closely as China’s advanced models become central to national technology policy.

  • Anthropic Accuses Alibaba of Massive AI Model Distillation Attack

    Anthropic Accuses Alibaba of Massive AI Model Distillation Attack

    Anthropic has publicly accused Alibaba of orchestrating a large-scale distillation campaign aimed at extracting capabilities from its Claude AI models. The allegation, detailed in a letter sent to U.S. senators, adds fresh tension to the ongoing technology rivalry between the United States and China.

    According to Anthropic, the e-commerce and technology giant used approximately 25,000 fraudulent accounts to generate over 28.8 million interactions with Claude between April 22 and June 5, 2026. The goal was to illicitly replicate the performance of Anthropic’s most advanced model, Claude Mythos Preview, using a technique known as knowledge distillation—a legitimate machine learning method that can be weaponized for model extraction attacks.

    This technique allows bad actors to feed input-output pairs from a proprietary “teacher” model into their own “student” model, effectively creating a cheap replica of the original system. Anthropic claims that Alibaba and its AI lab Qwen were behind the campaign, marking what it describes as the largest known instance of such an attack on the company.

    The accusation comes amid a rapidly closing frontier gap between Western and Chinese AI models. For example, Z.ai’s GLM-5.2 model, released shortly after Anthropic restricted global access to its most advanced model under U.S. government orders, has achieved benchmark performance nearly on par with leading Western frontier models. Z.ai has since captured a $128 billion market capitalization and plans to accelerate its pursuit of AGI.

    This is not the first time Anthropic has raised alarm over distillation attacks. Earlier in February, the company alleged that several Chinese AI firms—including DeepSeek, Moonshot AI, and MiniMax—had collectively generated millions of interactions with its Claude platform. Anthropic warned that such attacks are becoming more sophisticated and require closer coordination between AI companies and governments.

    The issue has also drawn attention in Washington. The Pentagon has added Alibaba to its list of Chinese military companies, a designation the company is contesting. Meanwhile, Reuters reported that the U.S. Commerce Department has so far held off adding DeepSeek to its trade blacklist, despite national security concerns, as officials weigh diplomatic repercussions.

    Alibaba has not yet responded to requests for comment on the allegations.