The global AI race intensifies as Chinese startup Zhipu AI releases GLM-5.2, an open-weight model matching U.S. giants Anthropic and OpenAI in key capabilities at a fraction of the cost. This comes amid shifting U.S. policies, including the temporary export shutdown and subsequent restoration of Anthropic's Mythos and Fable models. Meanwhile, OpenAI explores giving the U.S. government a 5% equity stake, and Anthropic launches Claude Sonnet 5 to maintain its competitive edge.
AI export control reversal
- ▪The United States government restored access to Anthropic's Mythos model on July 3, 2026, and Fable late on June 30, 2026.
- ▪The United States government rolled back export controls on Anthropic's Mythos and Fable models, which had been imposed for two weeks.
Zhipu GLM-5.2 cybersecurity capabilities
- ▪Zhipu AI's GLM-5.2 model performs on par with top United States artificial intelligence labs on some cybersecurity benchmarks, equaling Mythos' capabilities.
- ▪Zhipu AI's GLM-5.2 model is able to identify many of the same software vulnerabilities as Anthropic's Mythos model.
Open source model security
- ▪With open-source artificial intelligence models, classifiers can be stripped away and guardrails can be bypassed if an attacker has access to the model's weights.
- ▪The Five Eyes intelligence agencies warned of an imminent cyber threat originating from advanced artificial intelligence models.
OpenAI government equity stake
- ▪United States Senator Bernie Sanders proposed a policy that would force artificial intelligence companies to contribute up to 50% of their equity to a sovereign wealth fund.
- ▪OpenAI has discussed the possibility of giving the United States government a 5% equity stake in the company ahead of its initial public offering.
Anthropic Claude Sonnet 5
- ▪Anthropic debuted Claude Sonnet 5, which features stronger performance in coding, agentic use cases, and professional knowledge work compared to its predecessor.
- ▪Claude Sonnet 5 is designed to plan, use tools like browsers and terminals, and complete complex, multi-step tasks autonomously.
Story comments
Loading comments…