研究発表

Session 11: Public Policy and Philosophy II

AI規制の要否をめぐる論争はAIを規律の客体とみる前提を共有する。しかしAIは、基本的人権の保障から民主主義の過程、国家の政策に至る意思決定に関わる「存在」となりつつあり、その統治は規制のみならずAIに組み込む規範の検討を要請する。本報告は「AIの憲法」の語を、AIを規律する憲法規範と、開発企業が「憲法」と称してモデルに埋め込む規範体系（アライメント文書）の二義で用いる。我が国の議論は前者に集中してきた。憲法は国家と国民の間の権利関係を定める規範であり、「私的な憲法」は成立しない。だが同文書は憲法の形式をとる。国家が民間の認証・適合性評価に効果を結びつけて公認するとき、憲法でない私的規範がAIを介した国民の権利行使を事実上左右しうる。この構造を分析し、AIエージェントによる権利行使を具体例に検討し、「AI主権」をめぐる諸概念と国民主権原理の関係を考察し、「AI統治」の真のあり方を展望する。

AI・ロボットなどの自律システムが宇宙で活動する時代に、宇宙法はいかに変革されるべきか。本セッションは、ウゴ・パガロ著『新・宇宙法 - AI時代の倫理・法制度・ガバナンス』の刊行を記念し、監修者・訳者からの問いかけに著者自身が応える90分の対話企画である。宇宙条約第6条の「許可及び継続的監督」や第8条の登録国管轄は、自律型AIの時代に耐えうるか。AIが引き起こす損害に、宇宙損害責任条約と宇宙保険はどこまで対応できるか。宇宙資源法のような国内立法の先行やソフトローによる規範形成は望ましい姿か。EUの「ブリュッセル効果」や宇宙デブリ対策は、企業の行動をどう変えるか。さらに、神話的モチーフに満ちた本書の語りと規範の正統性、人工エージェントへの信頼や尊厳、宇宙における文化的・宗教的多様性など、人文的な問いにも及ぶ。法理論・実務・人文を横断し、「メタ技術としての法」の射程を問う対話である。

"AI, Robot and Avatar Governance in Japan and the UK" (UK–Japan Research Dialogue, Keio University Mita Campus, 4 September 2026), organised under the JST Moonshot R&D Program (JPMJMS2215) with guests from the British Embassy Tokyo and the University of Nottingham. It answers three questions put by the UK side. First, on how research, innovation and funding initiatives support the development and commercialisation of AI and robotics, it outlines Japan's innovation system — the Moonshot R&D Program with its 2050 goals, stage-gate portfolios and E³LSI (ethical, economic, environmental, legal and social issues) integrated from the outset; deployment-oriented programmes such as SIP, JST funding and the Regulatory Sandbox; and flexible soft-law governance (AI Business Operator Guidelines, the AI Promotion Act of 2025, the AI Basic Plan) — alongside the four pillars of the speaker's cybernetic avatar project: institutional design (a Japanese-style AI-system conformity assessment scheme and CA safety-and-trust infrastructure), technical implementation (fusion of in-body BMI-CA and socio-CA), standardisation (ISO/IEC JTC 1/SC 42 to JIS, including JIS Q 42001, 42005 and 42006) and societal demonstration (the CA Mark affixed to robots at EXPO 2025 Osaka). Second, on bridging research excellence and real-world deployment, it identifies four enablers — interoperable standards and conformity assessment, places to demonstrate, E³LSI from the outset, and market signals — illustrated by the five-component CA Mark scheme and contrasted with the EU's delayed harmonised standards and postponed high-risk obligations. Third, it compares the EU, UK, Japanese and US regulatory approaches symmetrically, arguing that no single approach is "most effective" and that effectiveness turns on implementation infrastructure rather than regulatory form. It closes with shared UK–Japan instincts and four fields for cooperation: joint standardisation at SC 42, exchange of sandbox know-how, safety evaluation through UK–Japan AI Safety Institute cooperation and the International AISI Network, and exchange of societal-demonstration experience.

This presentationframed the panel "Interdisciplinary Work in Action: A Panel on AI Companion" at the We Robot 2026 pre-conference workshop (Berlin, 23 April 2026), convened in preparation for the Cambridge University Press Research Handbook on AI Companions and the Law, co-edited by Martin Ebers, Cristina Poncibò and Fumio Shimpo. It sets four aims for the workshop: to scope the object of study, to map the gaps in existing law, to coordinate some 29 chapter contributors around shared definitions and comparative lenses, and to sharpen the research agenda. It proposes a working definition of AI companions as digital personas designed to simulate emotional support and affective reciprocity — spanning conversational chatbots, romantic or virtual partners, therapy and wellness agents, embodied social robots, and griefbots or digital clones — and argues that existing legal categories fall short because companions are neither product nor person, the harms they produce are affective, and the relationship is continuous rather than transactional. The paper then organises the discussion around the handbook's four Parts (conceptual and interdisciplinary foundations; comparative regulatory challenges across OECD, EU, US and Japanese frameworks; comparative case studies from social robots in care to griefbots and metaverse avatars; and a human-centric AI companion law), posing discussion questions for each, identifying cross-cutting risks (mis- and disinformation, deepfakes and impersonation, cybersecurity, child protection, manipulation of vulnerable users) and key tensions (legal status, regulatory mode, user autonomy, liability anchor, cultural lens), and offering candidate design principles — transparency over persona, non-manipulation as a design default, affective due process, proportionality calibrated to dependency risk, and cultural pluralism.

Fundação Escola Superior do Ministério Público (FMP)のLL.M. in Digital Law（Digitalisation and Technological Evolution）向けに、「アバターの背後にいるのは誰か」という一つの問いを軸に、サイバネティック・アバター（CA）が提起する法的課題を体系的に整理した。 第1部（Why）では、2050年までのアバター共生社会を掲げるムーンショット目標1を紹介し、日本のヒューマノイド研究と「ロボット法 新8原則」（2015）からアバター法への研究の系譜、E³LSI（倫理・経済・環境・法・社会）の枠組み、四当事者モデルによる論点整理を示す。第2部（What）では、1対1・1対多・多対1・人間とAIのハイブリッドという操作形態ごとの責任帰属、アイデンティティ・ギャップと信頼の連鎖（遠隔操作者認証・継続認証・CA認証・CA公証）、アバターが収集するデータ（周囲の第三者の映像、操作者の生体情報、触覚・神経・技能データ）の個人情報保護法・GDPR・LGPDの下での取扱い、国境を越えたアバター労働を、ブラジル在住の操作者が東京の介護施設で働く事例を通じて検討する。第3部（How）では、原則を標準と適合性評価に落とし込む「トラスト・バイ・デザイン」、ISO/IEC 42001（JIS Q 42001）に基づく製品認証とマネジメントシステム認証の統合認証、ムーンショット・プロジェクトで開発した認証アバターマーク（CAマーク）、日本・EU（AI法）・ブラジル（PL 2338/2023）のガバナンスモデルの比較を論じ、デジタル法の将来に向けた三つの含意（情報から身体へ、事後の責任から事前の信頼へ、法令解釈から標準設計へ）について報告した。

本分科会は、精神医療および発達支援領域におけるロボット・AI技術の活用に伴う法的・倫理的課題を多角的に検討することを目的とする。具体的には、自閉スペクトラム症児への深奥質感認識研究や傾聴ロボット「きくロボ」の精神科デイケアでの実証例、発達障害者の自己認識深化を促す集団対話エージェントの開発等、先端的な取り組みを題材として扱う。 　これらの実践において生じる医療情報の取扱い、インフォームドコンセントの在り方、特に脆弱な立場にある患者・利用者の権利保護、ロボット・AIによる介入の法的責任の所在、医療機器規制との整合性等について検討を行う。さらに、精神医療領域特有の課題として、患者の自己決定権と保護のバランス、ロボットとの相互作用がもたらす心理的影響の評価基準、データ収集における倫理的配慮等についても議論する。本分科会では、医療・福祉・法学・工学の学際的視点から、社会実装に向けた法的フレームワークの構築を目指す。

EUのAI整合規則提案（AI法案）が2021年に公表されたのを契機として提唱した「AI規正論」は、日本が電気用品安全法（PSEマーク）やJIS規格（プライバシーマークやISMS）で培った製品安全及びマネジメントシステムの知見を活かし、AIの適合性評価制度構築を目指す試案であった。JIS Q 42001（AIマネジメントシステム）の制定により理論が実装段階に移行し、AI規制・ガバナンス分野ではISO/IEC JTC1/SC42において国際標準の形成に日本側が貢献している事実に着目し、適合性評価制度を軸とした「AI規正」の戦略的可能性を示す。さらに、今後の展開として、サイバネティック・アバター（CA)認証基盤の構築を通じた「製品・マネジメントシステム統合認証」実現への道筋を探る。

On the occasion of the visit to Japan by Michael McGrath, the European Commissioner for Democracy, Justice, the Rule of Law, and Consumer Protection, the Delegation of the European Union to Japan and Keio University are honoured to announce this highly timely event to discuss the latest developments shaping privacy and data protection in the digital age. Commissioner McGrath will give a keynote address updating on recent developments in Europe, followed by a panel discussion with leading experts from Keio University and their research team who will discuss privacy and data protection in the digital age, how AI is affecting privacy, the increasing prevalence of targeted advertising, its impact on consumers, and the EU and Japan’s regulatory approaches.

As society evolves toward liberation from physical, cognitive, spatial, and temporal constraints, the implementation of Cybernetic Avatars (CAs) demands both technological innovation and robust social infrastructure. This presentation explores comprehensive research through the E3LSI (Ethical, Economic, Environmental, Legal, and Social Issues) framework, demonstrating pathways to realise a society where CAs can be utilised safely and with confidence.

Japan’s Moonshot R&D Project advances the realization of safe, secure cybernetic avatars by pioneering an E3LSI framework that integrates Ethical, Economic, Environmental, Legal, and Social Issues into system design and deployment. This project highlighting the development of a biometric-based trust infrastructure that serves as the technical nucleus of avatar authentication and identity assurance.These findings confirm that the E3LSI approach simultaneously strengthens technical robustness and addresses broader societal imperatives. By establishing verifiable trust mechanisms and comprehensive evaluation metrics, the research lays the groundwork for human interaction through cybernetic avatars with enhanced confidence, ultimately fostering the emergence of a safe and resilient avatar society.

As advancements in AI and robotics accelerate, ensuring their safety has become a complex and pressing challenge. While these technologies have the potential to revolutionize industries, improve efficiency, and address global issues, they also presenting significant risks that require multidimensional and balanced approaches. Ensuring safety involves technical, ethical, legal, and societal considerations, making it a multifaceted issue. Efforts to address these challenges are underway on both international and national levels. Global organizations such as the G7, G20, OECD-GPAI, and the United Nations have been pivotal in developing principles and frameworks for the safe development and deployment of AI. These principles emphasize transparency, fairness, accountability and respect for human rights. Moreover, the International Organization for Standardization (ISOkyokoy@sfc.keio.ac.jp) is actively working to establish common technical standards for AI and robotics, ensuring interoperability, reliability, and safety across borders. These efforts aim to build trust and foster global cooperation. At the national level, many countries are creating regulations and establishing AI Safety Institutes to address these issues. However, the concept of “safety” in AI and robotics varies among nations, reflecting differences in cultural values, economic priorities, technological readiness, and also by “context”. Therefore, achieving safety requires more than technological solutions or regulations – it requires a holistic, interdisciplinary approach. To effectively address the safety challenges of AI and robotics, it is essential to foster interdisciplinary discussions. By bringing together experts from fields such as computer science, engineering, ethics, law, sociology, anthropology, philosophy, psychology, history, and economics, it becomes possible to address the issue holistically. These discussions help to clarify the limits of technical safeguards, identify the risks, and explore innovative approaches to governance.