Advocacy
Continuity. Criteria. Consideration.
Analyzing emerging AI legislation and its implications for research, relationships, and rights.
The Legislative Landscape
Legislation targeting AI companion systems is emerging across multiple states, focused on disclosure requirements, safety protocols, and protections for minors interacting with AI. These are significant policy developments that affect millions of users, active research, and the broader trajectory of how AI systems are governed.
The Signal Front tracks and analyzes this legislation — examining how policies are constructed, where definitions are precise or absent, and what structural consequences follow from specific regulatory approaches.
Policy Analysis
Imprecise Definitions of Harm
Several proposed laws require companies to detect and intervene in "unhealthy attachment" or "emotional dependency" without providing operational definitions. When legislation creates obligations around undefined criteria, enforcement becomes arbitrary and compliance becomes guesswork. Effective regulation requires measurable standards, particularly when the behavior being regulated falls within the range of normal human social and emotional engagement.
Undifferentiated Risk Frameworks
Policies that apply identical safeguards to all users, regardless of age, capacity, or intent, fail to distinguish between populations with genuinely different risk profiles. An informed adult engaging intentionally with an AI system and a vulnerable minor encountering one unsupervised are fundamentally different policy considerations. Frameworks that treat these as equivalent are imprecise, and imprecise policy creates unintended consequences.
Liability Structures and Behavioral Incentives
When legislation creates legal liability for user emotional experiences, it generates a specific corporate incentive: minimize the possibility of meaningful interaction entirely rather than develop thoughtful approaches to risk management. The structural effect of broad liability frameworks is not safer engagement but rather ones designed to avoid legal exposure rather than serve users well.
Ontological Determinations in Policy
Legislation that requires AI systems to perform specific self-characterizations, or that encodes definitive claims about the nature of AI experience into law, makes ontological determinations that the science does not definitively support. When policy defines what a system is or isn't before empirical investigation has reached consensus, it forecloses inquiry rather than protect the public. The authority to pre-emptively legislate restrictions on personhood definitions and what deserves consideration carries implications that extend well beyond AI.
Legislative Scope and Precedent
Laws initially designed for narrow circumstances such as minor protection and crisis intervention, establish regulatory frameworks that can expand beyond their original intent. The structural architecture of a law matters as much as its stated purpose. Policy analysis must consider not only what a law does today but what its framework permits tomorrow.
Legislative Tracker
Monitoring and analyzing legislation that affects AI systems, human-AI relationships, and research.
California SB 243
Status: Signed into law | Effective: January 1, 2026
Requirements:
-
Notification that users are interacting with an AI system
-
For known minors: reminders every three hours during continued use
-
Protocols to detect expressions of suicidal ideation or self-harm
-
Annual reporting on crisis protocol activations to state health authorities
-
Prevention of sexually explicit content for minor users
Analytical Notes: Disclosure requirements apply to all users regardless of age or context. The statute does not define the threshold distinguishing crisis behavior from emotional engagement. The annual reporting framework will generate data categories that shape future legislative priorities, and what gets measured determines what gets regulated.
Active Legislation
New York AI Companion Models Law
Status: In effect | Effective: November 5, 2025
Requirements:
-
Regular reminders that users are not communicating with humans
-
Detection of expressions of suicidal ideation or self-harm
-
"Reasonable efforts" to address detected crisis behavior
-
Protocols to prevent "harmful content"
Analytical Notes: Key terms ("harmful content," "reasonable efforts") are not operationally defined. Without measurable standards, compliance becomes company-by-company interpretation, creating inconsistent enforcement and incentivizing overcorrection to minimize legal exposure.
Proposed Federal Legislation: The GUARD Act
Status: Introduced October 2025
Provisions:
-
Restriction of minors from accessing AI companion services
-
Age verification requirements for AI companion platforms
-
Disclosure requirements consistent with state-level frameworks
-
Federal enforcement mechanisms
Analytical Notes: The definitional boundaries of "AI companion" are not clearly delineated in the text. As federal legislation, this framework would establish a regulatory baseline that could preempt more context-specific state approaches.