Trump's Move Backfires as China's AI Now Far More Advanced
US President Donald Trump’s decision has once again become a ‘boomerang’. Recently, Trump ordered Anthropic to cut off access to its most advanced AI models, Mythos 5 and Fable 5, for foreign nationals, even if they are employees of the company. Anthropic responded by terminating access for all users. Fable is a large language model (LLM) based on Mythos, touted as the most advanced system currently available. Its capabilities have caused anxiety in many countries, prompting warnings to banks about its impact. Mythos is reportedly able to easily detect software vulnerabilities, even those missed by system developers. While beneficial, there are concerns that Mythos could be used by cybercriminals to launch more sophisticated and massive attack campaigns.
Not long after Anthropic’s most advanced AI models were completely blocked, China launched a new breakthrough. Just one day after the block, Chinese AI startup Z.ai (Zhipu AI) launched its flagship model, GLM-5.2. This move captured the attention of global users, as the GLM-5.2 performance benchmark trails only slightly behind leading closed-source models from the US, such as those from OpenAI and Anthropic. Z.ai revealed plans to use proceeds from domestic share sales to fund further AI development efforts, according to Reuters. This confirms earlier expert opinions that blocking Anthropic’s advanced Mythos model would not solve the problems and concerns arising from AI, as it would not be long before similar models are launched by other parties, including China.
‘Our mission is to achieve AGI (artificial general intelligence), so our current focus is on how to perfect our model to reach the highest level of intelligence. Therefore, all these resources are very helpful to us,’ said Qinkai Zheng, technical lead of the company’s CodeGeeX team. The accompanying surge in investor interest has driven the company’s share price up more than 2,000% since its spectacular debut in Hong Kong in January 2026. The company surpassed a market capitalisation of HK$1 trillion (Rp2,290 trillion) this week. ‘This model is comparable to top-tier closed models,’ Zheng told reporters at the company’s headquarters in Beijing. ‘This is the first time an open-source model has truly delivered extremely robust coding and agent performance, capable of competing with leading AI companies possessing exclusive technology like Anthropic and OpenAI,’ he stated. GLM-5.2 now ranks fourth on the Artificial Analysis LLM intelligence leaderboard and second on the Code Arena front-end coding leaderboard, with operational costs around one-sixth of those for leading US closed-source AI models.
The sudden termination of access to the leading US-made Anthropic AI model has sparked serious concern among global allies such as Canada and France. The leaders of both countries sharply criticised the excessive dependence on US-controlled AI infrastructure at the G7 summit last week. For the first time, a Chinese open-source AI model has nearly managed to close the technological gap with heavily funded Western AI labs, having previously surpassed US open-source models like Google’s Gemma and Meta’s Llama series, despite computing power limitations. Previously, analysts estimated that the performance capabilities of China’s top AI models lagged four to six months behind the leading US-made models.