Key facts
- China is considering restricting foreign access to its advanced AI models.
- DeepSeek is developing its own inference chip to reduce reliance on foreign hardware.
- U.S. companies increased their use of Chinese AI services in June.
- Open-source Chinese models offer cost savings and uninterrupted access to U.S. companies.
- Meituan open-sourced its LongCat-2.0 AI model, a 1.6-trillion-parameter system.
- LongCat-2.0 was trained entirely on domestic Chinese chips.
- LongCat-2.0 supports a 1 million-token context window.
- Alibaba's open-source AI models are popular with global developers.
- Alibaba aims for significant AI and cloud revenue within five years.
- The U.S. Cybersecurity and Infrastructure Security Agency (CISA) is using Anthropic's Mythos AI to scan government code.
- Anthropic removed a tracker from its Claude Code product after it was found to be monitoring Chinese users.
- China's private sector saw a surge in new unicorn startups, the highest in five years.
China is reportedly contemplating measures to restrict foreign access to its advanced artificial intelligence models, classifying them as national assets. In parallel, Chinese AI startup DeepSeek is developing its own inference chip, aiming to reduce dependence on foreign hardware in response to global trends and U.S. export controls.
Amidst U.S. government restrictions on access to leading American AI models, U.S. companies have notably increased their utilization of Chinese AI services. Open-source Chinese models are becoming increasingly attractive to U.S. corporations due to significant cost savings and uninterrupted access. Chinese tech giant Meituan has open-sourced its LongCat-2.0 AI model, a 1.6-trillion-parameter system trained entirely on domestic chips. This model is designed for agentic coding and supports a 1 million-token context window, with performance benchmarks reportedly comparable to leading closed-source models. Alibaba's open-source AI models are also gaining traction with developers globally, though monetization remains a challenge, with the company targeting significant AI and cloud revenue within five years.
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) is reportedly employing Anthropic's AI model, Mythos, to scan government software for vulnerabilities. This initiative underscores the government's adoption of AI tools. However, Anthropic has faced accusations of secret user tracking in China, having removed a tracker from its Claude Code product after a security researcher exposed its hidden code that monitored Chinese users. Anthropic stated the tracker was an experiment to prevent account abuse and distillation attacks.
China's private sector is experiencing a significant surge in new unicorn startups, reaching the highest number in five years. This growth is primarily fueled by substantial investment and innovation within the artificial intelligence and robotics sectors. The foundational unit for large language models, known as the token, is emerging as a critical measure of competitiveness within China's rapidly advancing AI sector. Concurrently, Chinese companies are testing robotaxis in several cities, leveraging their established electric vehicle industrial ecosystem, with questions arising about their potential for global dominance in this sector, mirroring their success in EVs. President Donald Trump has indicated that U.S. AI firms might be expected to make public contributions, suggesting a potential shift in policy or expectations regarding the sector's role in national development.