Weekly AI News Update (28-04 December 2025)

Weekly AI News Update (28-04 December 2025)

Tsinghua University has established China’s first comprehensive institutional framework governing AI use in education. The “Guiding Principles” define AI as a secondary tool, mandate human responsibility, and strictly prohibit its use for academic misconduct such as plagiarism or ghostwriting in coursework and research. 🔗 ContentGrip
Washington State is partnering with Microsoft to integrate AI into classrooms statewide. The initiative provides Microsoft’s Co-Pilot tool to all students at no cost, trains 2,000 teachers, and is supported by new state guides for the safe, ethical, and human-centred use of the technology. 🔗 NBC Right Now
OpenAI has introduced a new “confession” framework for training AI models. This approach incentivises models to acknowledge and honestly report undesirable behaviours, such as sycophancy or test hacking, separate from judgements of the main answer’s accuracy or helpfulness. 🔗 Engadget
Sam Altman, CEO of OpenAI, has declared an internal ‘code red’ to urgently refocus development on improving ChatGPT’s core performance and reliability. This strategic shift follows competitive pressure from rivals like Google, whose Gemini 3 model has demonstrated leading benchmark results, compelling OpenAI to delay several non-core initiatives. 🔗 The Verge
Kuaishou has launched its Kling O1 AI model, a unified multimodal system for video generation, editing, and understanding. The company positions it as a competitor to tools like OpenAI’s Sora and cites its integrated editing capabilities as a key feature for professional creative industries. 🔗 South China Morning Post