← 返回研究新聞
A diverse youth support team reviews multilingual AI conversations, curated knowledge and risk-escalation pathways with student representatives
政策 / 倫理產業訊號20262026年7月11日· 8 min

Product news: UNESCO IITE seeks a shared five-language youth AI assistant with risk-aware prompt optimization

UNESCO Institute for Information Technologies in Education

AI Product and Learning Report

500 字摘要

A diverse youth support team reviews multilingual AI conversations, curated knowledge and risk-escalation pathways with student representatives

UNESCO's Institute for Information Technologies in Education published a call for proposals on July 10 for a new multilingual AI-enabled chatbot and the consolidation of two existing services. Proposals were due on July 27. The notice is a procurement specification, not a product launch or contract-award announcement. It nevertheless provides an unusually concrete public description of the intended system, delivery responsibilities and responsible-AI checks for a service aimed at students and young people in Eastern Europe and Central Asia.

One workstream would consolidate the Oilo and Aspan chatbots on a shared Telegram solution supporting Russian, Kyrgyz, Kazakh, Turkmen and Uzbek. The services would retain distinct identities and functions while drawing on a common technical core, curated knowledge and safety logic. A separate AI Assistant for students and young people would operate through VK and/or Telegram. The specification says the tools should remain free for end users. An earlier UNESCO IITE account described Oilo as using expert-curated information on health, mental health, relationships and safety.

The proposed Prompt Optimization Tool is designed to clarify or reformulate a user's query in real time. It should identify the topic and level of risk, then help produce a safer, more structured and evidence-informed response. This intermediate step could be valuable when a young person's wording is brief, ambiguous or emotionally charged. It is also a consequential classifier: a missed risk or poorly reframed question could send the conversation in the wrong direction. The public notice does not report measured accuracy for topic detection, reformulation or escalation.

The requested service includes high-risk response protocols, testing, correction of critical defects, technical documentation and source-code handover. The contractor would also address relevant sections of UNESCO's Ethical Impact Assessment 2.0 and provide hosting, model access, APIs and maintenance through the end of 2026. These requirements make governance part of the build rather than an afterthought. They do not specify which bidder, model or final architecture would be selected, and a written safety requirement is not evidence that the deployed system will satisfy it.

The boundaries should remain prominent in publication. No contract award, completed assistant, user study, safety audit, uptime result or learning outcome is established by the call. Telegram and VK access varies by jurisdiction and institutional policy. Curated information can become outdated, multilingual meaning can shift, and a youth-facing chatbot cannot replace professional health, mental-health or safeguarding services. Reporting should therefore say UNESCO IITE invited technical proposals, not that UNESCO had already launched a five-language assistant.

For Hong Kong, the specification offers a governance pattern rather than a ready product. A local youth service would need Traditional Chinese, Cantonese and English evaluation, locally verified support information, Personal Data (Privacy) Ordinance review, age-appropriate consent and clear handoff to trusted adults or emergency resources. Procurement should test false negatives and false positives in risk classification and let domain experts review reformulations. The educational value lies in making curated evidence, multilingual quality, risk routing and accountable human support visible requirements before development begins.

相關論文

大學生向教師解釋幾何作圖,同學在旁思考,桌上的電腦展示相關數碼圖解
政策 / 倫理2026年9月7日
政策 / 倫理 112

評論:Astra 的 AGI 主張,讓教育更需要看見人的真實學習

AIED.HK Editorial

AI Product News Commentary

OpenAI 於 2026 年 9 月 3 日發布 GPT-6 Astra,並引發關於 AGI 是否已經到來的討論。本文將這一說法視為需要歸屬的主張,而非已確立的共識。教育眼前的挑戰,是分清 AI 能產出甚麼,以及學習者能獨立解釋、質疑和遷移甚麼,再以更強的代理能力支援真正的學習。

產品新聞評論GPT-6 Astra
閱讀 500 字摘要 →
三位教育與軟件同事在明亮的大學設計工作室審查圖解教材卡、註釋圖表和數碼原型
政策 / 倫理2026年9月7日
政策 / 倫理 113

評論:Fable 5.1 把更長程的 AI 工作帶進 AIED,教育驗證更顯重要

AIED.HK Editorial

AI Product News Commentary

Anthropic 於 2026 年 9 月 1 日發布 Claude Fable 5.1,強化長程編程與知識工作能力,並降低快取讀取價格。AIED 的機會,是加快從教學構想到可審查原型與研究分析的循環;真正的考驗,是能否把速度轉化為更好的教學與可信證據,同時計入總成本、資料條件和人工審核。

產品新聞評論Claude Fable 5.1
閱讀 500 字摘要 →
A lecturer and two university students inspect ranked learning tools, separate cloud and local plugin cards, and a review ledger in a bright computing studio
政策 / 倫理2026年8月23日
政策 / 倫理 111

Product news: ChatGPT plugin ranking and Claude Code 2.1.239 make tool selection and workspace boundaries inspectable

OpenAI, Anthropic, Google for Education

AI Product and Learning Report

Product news: ChatGPT now ranks plugin recommendations partly by continued use after installation and adds more time-aware answers, while Claude Code 2.1.239 distinguishes cloud-synced plugins from local installations and makes a data-residency cost premium visible. Gemini for Education supplies the institutional purpose boundary across teaching, learning and work. Together, the updates make tool selection, context, cost and human review part of AI workflow literacy.

product newsChatGPT pluginsClaude Code 2.1.239
閱讀 500 字摘要 →