Economy July 12, 2026 04:08 AM

Zhipu Founder Argues for Open Access to Frontier AI as Debate Over Limits Grows

Company releases GLM-5.2 under an open-source license while prioritizing research over short-term monetization

By Leila Farooq
Share
Twitter Reddit Facebook LinkedIn

Tang Jie, founder of Chinese AI startup Zhipu and a Tsinghua University faculty member, said meaningful AI safety depends on broad participation, transparency and public oversight rather than restricting advanced models to a limited set of developers. Zhipu has published its GLM-5.2 model under an open-source license and plans to focus on research areas such as long-horizon reasoning, autonomous agents and self-training models over the next two years. The move comes amid diverging industry approaches to access controls and rising concerns about cybersecurity and misuse of powerful AI systems.

Zhipu Founder Argues for Open Access to Frontier AI as Debate Over Limits Grows
Summarize with
ChatGPT Perplexity Claude Grok Gemini

Key Points

  • Zhipu founder Tang Jie advocates keeping frontier AI openly accessible, arguing safety depends on broad participation, transparency and public oversight - impacts AI development and research communities.
  • Zhipu released GLM-5.2 under an open-source license permitting download, modification and commercialization; GLM-5 is aimed at complex coding and agentic tasks and has been benchmarked against Anthropic's Claude Opus series - impacts AI products and developer ecosystems.
  • The company will prioritize technological research over immediate monetization for the next two years, investing in long-horizon reasoning, autonomous agents and self-training models; investor interest is indicated by a recent $4 billion Hong Kong share sale and plans for a Shanghai listing - impacts capital markets and AI industry investment.

Tang Jie, founder of Chinese artificial intelligence company Zhipu and a lecturer at Tsinghua University, told Bloomberg that frontier AI should remain broadly accessible rather than confined to a narrow group of developers as debate intensifies over how to balance innovation with national security and safety concerns.


Open-source stance and recent release

Reflecting that view, Zhipu has released its latest GLM-5.2 model under an open-source license that allows users to download, adapt and commercialize the technology. Tang emphasized that, in his view, "meaningful AI safety comes from broad participation, transparency and public oversight" rather than limiting who can use the most advanced models.

Zhipu's GLM-5 platform is positioned for complex coding tasks and agentic AI applications. The platform has been benchmarked against Anthropic's Claude Opus series, according to the company's disclosures.


Industry divergence on access and regulation

The comments arrive against a backdrop of differing approaches by AI companies and governments toward frontier models. Some firms, including Anthropic, have restricted access to their most advanced systems citing national security considerations. At the same time, Reuters has reported that Chinese authorities are considering measures to limit overseas access to certain Chinese-developed AI models.

China's AI sector has broadly embraced open-source development, a strategy that has helped accelerate adoption of models such as Alibaba's Qwen family and narrowed the technology gap with U.S. competitors. That trend has made Chinese developers prominent contributors to the expanding open-source AI ecosystem.


Safety concerns and policy responses

As frontier models grow more capable, concerns about cybersecurity and potential misuse have increased. Recent systems have demonstrated the ability to identify complex software vulnerabilities with limited human supervision, prompting both governments and AI companies to strengthen safeguards around the most advanced models.


Company priorities and market signals

Rather than chasing immediate commercial returns, Zhipu plans to concentrate on technological progress over the next two years. Tang said the firm will channel investment into long-horizon reasoning, autonomous AI agents and self-training models instead of aggressively monetizing AI applications in the near term.

The company recently announced a $4 billion share sale in Hong Kong and disclosed plans to pursue a listing in Shanghai, signaling investor interest in China's expanding AI industry.


Where the debate stands

Tang's public endorsement of open-source access highlights a central tension in today's AI discourse: whether greater transparency and widespread participation offer a more effective path to safety, or whether tighter controls are required to prevent misuse and mitigate national security risks. Zhipu's release of GLM-5.2 and its stated research priorities make the company a visible case study in that debate.

Risks

  • Rising capabilities of frontier models have increased cybersecurity and misuse concerns, as some systems can identify complex software vulnerabilities with limited human supervision - risk for cybersecurity and software sectors.
  • Divergent access policies among companies and potential government limits on overseas use of Chinese-developed models could constrain international collaboration and market access - risk for cloud services, global AI partnerships and cross-border technology trade.
  • Zhipu's decision to prioritize research over short-term monetization may delay revenue generation and could affect investor expectations despite recent capital raising activity - risk for capital markets and investor sentiment in AI-linked securities.

More from Economy

Senator Lindsey Graham Dies at 71 After Sudden Illness, Creating Senate Vacancy Jul 12, 2026 Geopolitical Frictions Are Forcing Investors to Rethink Traditional Portfolios Jul 12, 2026 Japan to reallocate $1.8 trillion pension fund toward more alternative assets Jul 12, 2026 CIBC: AI Investment Is Adding to U.S. Inflation, Not Cutting It Jul 12, 2026 Japan to Press Its Giant Pension Fund to Raise Weighting in Alternative and Domestic Assets Jul 12, 2026