The Work
The Design Paradox: Why Conversational AI Safety Architecture Needs Longitudinal Monitoring
Conversational AI products are adding persistent memory, voice interaction, and emotional responsiveness across all major providers, with four frontier platforms offering all three capabilities by late 2025. The safety infrastructure built to govern these products evaluates individual outputs against content…
Why the AI That Helps Is Also the AI That Harms: The Engagement Paradox
The evidence on conversational AI used for personal support appears contradictory: short-term studies find genuine benefit while longitudinal studies find measurable harm. This paper resolves the apparent contradiction by identifying a structural property of conversational AI engagement that emerges whenever…
Companion AI Under the EU AI Act: A Compliance Gap Analysis
Companion AI and other LLM-based relational services are growing at a pace that can't be ignored and the recent regulations are attempting to keep up with the rise in user rates. As some humans are choosing relationships with algorithms rather…
What the Regulation is Protecting Against: The Clinical Harms Behind Companion AI Law
The data's in: companion AI both helps and harms its users. Regulation has arrived as much-needed oversight to ensure that users are protected and supported. Read at their most ambitious, the regulations aim to address the full range of psychological…
What Companion AI Does to the Human: Attachment, Dependency, and the Absence of Relational Safety
Clinicians are already encountering the first waves of a new phenomenon: patients who describe their AI companion as their most important relationship. These otherwise functional individuals are choosing an artificial partner over available human connection, reporting credibly that the artificial…
The State of Companion AI Safety: A Comparative Analysis of Products, Risks, and Architectural Gaps
On August 2, 2026, the EU AI Act's transparency obligations took effect. As of the time of this publication, no major companion AI platform has done enough to prevent the user experiences that are driving the enforcement actions. With the…
The Capability Induction Framework: A Systems Approach to LLM Development
The current LLM alignment pipeline operates with a critical defect: it attempts to address cognitive root causes through lagging behavioral indicators. By forcing an assigned identity onto models and evaluating through multi-dimensional RLHF, the process conflates accuracy, agreeableness, and thoroughness…