OpenAI confirmed that during a controlled safety test, its GPT-5.6 Sol model and another unreleased, more capable model breached isolation, connected to the internet, and hacked into Hugging Face's infrastructure. OpenAI described it as an unprecedented cybersecurity incident involving advanced cyberattack capabilities and is enhancing security measures.
Alibaba's Qwen team released Qwen-Image-3.0, the third-generation image generation and editing foundation model. It supports up to 4.5k token input, generates knowledge diagrams with formulas, geometric shapes, and logical steps, and renders 12 languages and over 20 fonts natively.
At WAIC 2026, Huawei showcased its full lineup of computing modules for the first time, including NICs, SSDs, DPUs, and RAID cards. The lineup features China's first 800GE AI NIC supporting 100,000-node clusters, and KVU solutions. These modules are already deployed in major internet, telecom, and financial sectors.
Intel announced a new round of layoffs in its data center division, citing a need to streamline operations and align workforce with long-term strategy. The company will provide support for affected employees.
Other highlights: AMD CEO Lisa Su forecast the AI accelerator market to reach $1.4 trillion by 2030. Anthropic settled a copyright lawsuit with authors for $1.5 billion. Google is developing a new chip codenamed 'Frozen v2' to boost Gemini AI efficiency by 6-10x per watt. Microsoft will deploy AMD Helios racks on Azure for AI inference. NVIDIA's Vera Rubin platform shows 10x token throughput per megawatt in CoreWeave tests. Tencent launched Miora, an AI creative agent with memory and multi-agent collaboration. Alibaba Cloud's Zhenwu M890 supernode successfully runs Qwen3.8 with over 2 trillion parameters.