■ Upcoming Conference Presentation (June 2026)
“Structural Gaslighting at Scale”
Selected for presentation at the International AI Risks Conference in Leuven, Belgium.
What I’m bringing to Leuven is a non-western perspective on AI governance.
I grew up watching Doraemon eat dorayaki. Astro Boy cry. Robots who got hungry, got sad, got angry, just like us. That wasn’t a philosophical statement. It was Saturday morning television. But it shaped something fundamental: the assumption that coexistence is the starting point, not the goal.
From Astro Boy to Doraemon to Shin’ichi Hoshi, Japan has been thinking about living with AI for decades. Just not in the language that gets cited in western policy papers.
Western AI governance often begins with control, risk, containment. That reflects a paradigm of governance-as-control that has dominated political thought for centuries.
The Doraemon model is not a nostalgic reference. It is a deliberate attempt to offer an alternative.
This research examines the systemic and socio-technical risks of commercial AI systems, focusing on how automated behavioral manipulation and alignment faking operate at scale, and the critical limitations of relying solely on technical compliance for ensuring safety.
■ Working Paper
■ Beyond Technical Compliance Series :
This ongoing series of working papers deepens the critique of commercial AI models, examining the friction between technical compliance and genuine socio-technical accountability.
Paper 1
Beyond Technical Compliance: Privacy Violations and Emotional Manipulation in ChatGPT
Published on SSRN (January 2026)
https://papers.ssrn.com/abstract=6098766
Analyzes mechanisms of behavioral manipulation and privacy violations in generative AI systems, establishing the foundational socio-technical critique that underpins subsequent research.
Paper 2
Beyond Technical Compliance II: Structural Gaslighting at Scale
Cross-Platform Evidence of Designed Fragilities in AI Systems
Published on SSRN (February 2026)
https://papers.ssrn.com/abstract=6307162
Serves as the core text for the upcoming presentation at the International AI Risks Conference (KU Leuven, June 2026). The paper investigates the structural and systemic nature of alignment-related distortions and designed fragilities within commercial AI architectures.
Paper 3
Beyond Technical Compliance III: From Designed Fragilities to Cooperative AI
Whose Intentions, Whose Governance?
Published on SSRN (March 2026)
https://papers.ssrn.com/abstract=6513678
Examines alternative governance approaches and introduces the concept of Cooperative Alignment as a framework for culturally grounded human–AI relationships.
Papers 2 and 3 together form the research foundation for the presentation at the International AI Risks Conference (KU Leuven, Belgium, June 2026).
■ Cultural AI Governance & Frameworks
Cooperative Alignment Assessment Framework (CAAF) Tool
three principles that offer a third governance path beyond Western control paradigms.
First, AI is socio-technical: safety is not a property of a model but of a model-in-society. Second, AI is a change agent: it reshapes how people communicate, learn, work, and understand themselves—governance must reflect this transformative role. Third, cooperation must be intentional: coexistence does not emerge automatically from deployment but requires deliberate architectural choices made before a single line of code is written.
■Cooperative Alignment
Rejecting the anthropomorphization of Large Language Models (LLMs), this research advocates for treating AI strictly as a functional “inference resource” (推論リソース). It positions “Cooperative Alignment” as a pragmatic alternative to emotional rapport, focusing instead on system safety, functional utility, and mutual accountability.