Weekly AI News Update (12-18 September 2025)

Engineers from DeepSeek have detailed their reinforcement learning approach for training the DeepSeek-R1 reasoning model, published in Nature, achieving superior performance in mathematics and coding while addressing training efficiency challenges. South China Morning Post
Experts highlight AI's rapid integration into higher education, urging students to assess universities' strategies, develop AI literacy across disciplines, and prepare for evolving labour markets where adaptability and interdisciplinary skills remain critical. The Guardian
Hong Kong's Policy Address highlights education reforms, pledging HK$2 billion to embed AI in school curricula and expanding nonlocal student intake to strengthen the city's global talent and education hub role. China Daily
In the United States, educators report AI use is blurring definitions of cheating. Schools are shifting from bans to guidelines, emphasising AI literacy and clearer academic integrity policies. CGTN
Beijing has introduced AI courses in over 1,400 primary and secondary schools, offering at least eight classes annually to enhance students’ AI literacy, creativity, and problem-solving skills through practical, context-specific projects. Xinhua