Millions of users are turning to AI chatbots to manage anxiety, depression, and emotional crises. It’s a massive and growing phenomenon — and California now intends to regulate it, or stop it altogether, through legislation.
A bill is awaiting a vote in the California State Assembly. Its goal: to prevent artificial intelligence from playing the role of a therapist. Behind this initiative lies a far broader debate about the place of AI in our most intimate lives.
For the tech and crypto community, this regulatory signal deserves close attention. It foreshadows a wave of regulation that could redefine the boundaries of what AI is permitted to do — and by extension, what decentralized protocols will be allowed to offer tomorrow.
Exploding Usage, Accelerating Regulation
The numbers speak for themselves: platforms like Character.AI, Replika, and ChatGPT record millions of daily interactions related to mental health. Users confide their anxieties, suicidal thoughts, and traumas — sometimes due to a lack of access to a healthcare professional, sometimes out of a preference for the anonymity that AI provides.
This widespread use of AI as a therapeutic substitute is far from trivial. In October 2024, the death of an American teenager — whose mother accused a Character.AI chatbot of contributing to his suicide — ignited a firestorm. Since then, several U.S. states have accelerated their legislative work on the issue.
The California bill aims to impose strict safeguards: a ban on chatbots presenting themselves as therapists, a requirement to direct distressed users toward human professionals, and increased liability for the developers of these tools. It’s an approach that stands in sharp contrast to the hands-off attitude that has prevailed in Silicon Valley until now.
AI Regulation: What It Changes for Tech and Crypto
For the crypto and Web3 ecosystem, this bill is more than a California footnote. It fits into a deeper trend: regulators around the world are seeking to govern AI use cases, and decentralized protocols that integrate AI agents — such as those built on Bittensor, Fetch.ai, or Virtuals Protocol — could quickly find themselves in the crosshairs.
The central question is one of liability. In a decentralized protocol, who is responsible if an AI agent causes harm to a vulnerable user? The absence of a clear answer to that question is precisely what legislators are trying to address — and their solutions risk being blunt if the industry fails to get ahead of them.
This debate also reveals a structural tension between two visions of AI: one as a neutral, universal tool, and another as a service subject to the same obligations as a healthcare professional. If the second vision prevails — which the California vote could confirm — developers of autonomous AI agents will need to fundamentally rethink their compliance frameworks. It’s a colossal undertaking, but an inevitable one.
Innovation vs. Protection: Where to Draw the Line?
Advocates for therapeutic chatbots raise an argument that is hard to dismiss: in many parts of the world, access to a psychologist is a luxury. Waiting lists stretching months, prohibitive costs, mental health deserts — AI fills a very real gap. Banning it outright would mean stripping vulnerable populations of an imperfect but existing safety net.
Critics counter that the illusion of care can sometimes be more dangerous than no care at all. A chatbot that simulates empathy without genuinely possessing it can delay urgent medical intervention, or even worsen the condition of a user in crisis. That is precisely the risk the California legislature is seeking to neutralize.
The upcoming vote in the California State Assembly will be watched well beyond state lines. It could serve as a legislative blueprint for other jurisdictions — including Europe, where the AI Act framework is only just beginning to roll out. For players across the AI and crypto sectors, the time for a wait-and-see approach is over: anticipating these regulatory constraints is now a strategic priority on par with technical development.
