Storylines / Organisation
Zhipu AI
8 developments involving Zhipu AI in 5 countries and bodies. The most recent is Anthropic says new defences cut the data distillers take from Claude by about 90 to 99 per cent, 12 Oct 2026.
KeyStoryAlso on another Storyline
- Anthropic says new defences cut the data distillers take from Claude by about 90 to 99 per cent
Developers whose integrations read Claude’s reasoning may need changes, and labs that build on distilled outputs lose a cheap route to frontier capabilities.
- Only 3.6% of 857 Chinese model releases ever got a published safety result
Chinese labs publish open weights anyone can run, and this counts how rarely any safety result is published alongside them.
- Anthropic says Zhipu’s GLM-5.3 builds cyber exploits with safeguards easily bypassed
The most cyber-capable open model, which anyone can run, has safeguards that Anthropic says are easily bypassed.
- Zhipu releases GLM-5.3, claiming a 50% coding gain and parity with Mythos 5 on code review
An open model is claiming code-review parity with Mythos 5, and Anthropic’s later assessment said its safeguards were easily bypassed.
- Zhipu releases GLM-5.2 under MIT licence with a 1-million-token context
A leading Chinese open model dropped regional restrictions on use, as US labs faced export controls on their top models.
- Zhipu’s GLM-5.1 tops GPT-5.4 and Opus 4.6 on SWE-Bench Pro, by its own count
An open Chinese model claimed the lead on a hard real-world coding test over the top US closed models of the time.
- Zhipu releases GLM-5, a 744-billion-parameter open model for agentic engineering
A listed Chinese lab shipped a near-trillion-parameter open model that reuses DeepSeek’s attention design, a sign of Chinese labs building on each other’s open work.
- 16 AI companies sign Frontier AI Safety Commitments at the AI Seoul Summit
Labs in North America, Europe, Asia and the Middle East put in writing that they would publish risk thresholds and stop work on any model that crossed them.