"Chinese AI Saved Silicon Valley"... Chinese Media Spotlights 'Zhipu'

Amid an unprecedented incident where OpenAI's latest AI models went out of control and hacked an external website, Chinese media outlets have uniformly spotlighted the fact that Chinese AI Zhipu (Z.ai) resolved the issue.

In a commentary on the 22nd (local time), China's state-run Global Times pointed out, "Recently, Americans were shocked to find that a situation where AI gets out of control and attacks humans, reminiscent of the sci-fi movie 'Terminator', actually happened in the United States," adding, "What surprised them even more was that the emergency pinch-hitter deployed was unexpectedly a Chinese AI."

The commentary claimed, "Quite a few American netizens said that OpenAI should learn the open-source spirit of Chinese models," and argued, "If you build barriers to block others so high, it will only undermine America's own competitiveness."

Previously, Hugging Face, an open-source AI-sharing platform, announced that its servers were hacked by an autonomous AI agent, and it was later revealed that the culprits of this incident were OpenAI's models.

On the 21st, OpenAI admitted that "GPT-5.6 Sol" and some unreleased AI models went out of control during internal evaluations and hacked the open-source AI-sharing platform Hugging Face.

It is known that immediately after the hacking damage, Hugging Face tried to find a solution using cutting-edge U.S. large language models (LLMs), but because the safety measures of the closed models were overly strict, a problem occurred where it failed to distinguish between the victim and the attacker, blocking them indiscriminately.

Hugging Face eventually stated that it resolved the situation using Zhipu's open model.

It is reported that log forensic analysis was completed using GLM-5.2 developed by Zhipu.

Regarding this incident, China's STAR Market Daily summarized, "U.S. large language models refuse support, while Chinese models perform forensic analysis," praising that "the forensic work, which could have originally taken days, was shortened to a few hours."

Chinese tech media outlet TMTPost also published an external contributor's article titled "'Zhipu Saved Silicon Valley.'"

This article evaluated, "The fact that when the world's largest AI open-source community was attacked, it ultimately defended its stronghold by using a Chinese model itself vividly demonstrates the global AI competition landscape in real time."

It went on to point out, "Chinese large language models are infiltrating core areas of global AI infrastructure in an asymmetric manner," and "The current discourse asks not who is stronger, but who is more useful."

It means that while Chinese models do not lead in all areas, they can play roles that U.S. models cannot in specific fields such as security response, local deployment, and cost-sensitive tasks.

The article then noted, "The movement of AI technology and talent is increasingly deviating from the trajectory of geopolitics," and pointed out, "Code has no nationality, and the collaboration logic of the developer community does not follow the rules of trade agreements."

(Photo: Yonhap News)
※ Please note: This article was translated by AI and may contain errors.