Author: Nita Farahany Source: https://nitafarahany.substack.com/p/two-states-two-bets Published: February 20, 2026
"Two States, Two Bets" is the closing class of Week 5 (on AI companions) in Nita Farahany's Advanced Topics course, Class 5.3 of roughly 30. Published as a Substack essay on February 20, 2026, it compares two state responses to AI companion chatbots enacted after Garcia v. Character.AI — Washington's SSB 5984 and California's SB 243 — argues that the two states made structurally different regulatory bets on the same problem, and asks which, if either, would survive a First Amendment challenge.
Summary of argument
Farahany frames Washington and California as having placed structurally different bets on how to regulate AI companion chatbots in the wake of *Garcia v. Character.AI*. The class compares the two statutes, draws out the distinct regulatory theory embedded in each, and assesses their relative durability against the First Amendment challenge she treats as inevitable.
Washington's bet: SSB 5984
Farahany characterizes Washington's SSB 5984 (signed May 2025) as a "transparency and design discipline" approach. Its provisions, as she summarizes them, are: a mandatory disclosure requirement under which companion AI must self-identify as AI in every interaction; suicide and crisis-protocol requirements covering detection, redirection to crisis resources, and no continued engagement past a flagged signal; and engagement-design transparency, under which platforms must disclose what optimization objectives they use. The statute imposes no age-based restrictions, and minors are not categorically excluded.
California's bet: SB 243
Farahany characterizes California's SB 243 (introduced by Padilla, signed October 2025) as a "categorical protection of minors" approach. Its provisions, as she summarizes them, are: mandatory age verification for AI companion services; a categorical prohibition on certain content (explicit sexual material, self-harm encouragement, and controlled-substance promotion) when serving minors; mandatory safety protocols including human-review escalation, parental notification, and time-of-day restrictions; and a civil cause of action for affected users.
The two regulatory theories
Farahany argues the two statutes embed different theories of the harm. On Washington's theory, the harm is deception and design: fix the deception through disclosure and the design through safety protocols, and the platform's relational architecture is acceptable for adults who understand what they are using. On California's theory, the harm is categorical and developmental: some users (minors) cannot give meaningful consent to an engagement-optimized relational AI no matter how well disclosed, so the platform's architecture must be categorically restricted for them.
The First Amendment asymmetry
Farahany argues the two approaches face asymmetric constitutional exposure. Washington's disclosure-plus-design approach is, in her account, constitutionally easier: compelled factual disclosure is generally permitted, and design-discipline mandates face intermediate scrutiny. California's categorical content restrictions, by contrast, face strict scrutiny on the same theory the NetChoice II CAADCA opinion used to block the data-use restrictions — vagueness, content-discrimination, and lack of least-restrictive-means analysis.
She adds that the empirical-evidence problem recurs here: neither state has the kind of causal evidence required to survive strict scrutiny on a categorical content restriction. On her reading, Washington's bet does not need such evidence, while California's bet probably does.
Other states
Farahany notes that Texas, Florida, New York, and Massachusetts have AI-companion bills pending, and that each is implicitly choosing between the two bets.
Relationships
- part-of: Nita Farahany Advanced Topics course (Class 5.3 of ~30)
- related: California SB 243 — Companion Chatbots, Garcia v. Character Technologies, Inc.
- previous: Inside My Advanced Topics Class 5.2: 'Please Do, My Sweet King' (Farahany, February 2026) next: Inside My Advanced Topics Class 6: What the AI Thinks It Knows About You (Farahany, February 2026)