While Washington debates a national framework for artificial intelligence, the most consequential AI regulation in the United States is being written in state legislatures — and its focus is protecting children. According to the Transparency Coalition's 2026 mid-year tally, 14 chatbot-safety measures have been passed or enacted across 13 states this year, making conversational AI the fastest-moving corner of AI law in the country.
A Bipartisan Wave
What distinguishes this legislative surge is its striking political breadth. Chatbot-safety laws have been enacted in Republican-held legislatures such as Georgia and Idaho and in Democratic strongholds such as Washington and New York, often advancing with near-unanimous votes. In a policy arena defined elsewhere by ideological gridlock, lawmakers across the spectrum have found common ground in a shared concern for minors.
That consensus has shaped the content of the laws as much as their passage. Most focus narrowly on transparency and crisis response rather than attempting to regulate AI systems broadly — an approach that makes the bills easier to pass and harder to challenge on First Amendment or federal-preemption grounds. The Future of Privacy Forum, which tracks the space, counted 98 chatbot-specific bills across 34 states earlier this year, a pipeline that all but guarantees the count will keep climbing.
What the Laws Require
Despite differing sponsors, the enacted laws share a common structure. In broad terms, businesses operating conversational AI must:
- Disclose clearly that a user is interacting with an AI rather than a human.
- Implement safeguards for minors, including age-appropriate protections.
- Detect and respond to expressions of self-harm or suicidal intent, typically by surfacing crisis resources.
The newest frontier targets companion chatbots specifically. Idaho, Oregon, and Washington have enacted laws barring such systems from claiming to be sentient or initiating sexual conversations with minors — a direct response to the rise of AI "companion" apps marketed for emotional connection.
Two statutes stand out. California's SB 243, signed by Governor Gavin Newsom, combines user disclosure, crisis-intervention protocols, and minor-focused safeguards in a single framework, and is among the first laws to tailor protections specifically to children. New York's S-3008C requires any AI companion offered in the state to maintain protocols for detecting and responding to suicidal behavior and to clearly identify itself as non-human. Separately, Wyoming's HB 102, effective July 1, 2026, targets deepfake sexual depictions and the promotion of self-harm to minors, with penalties that reach felony charges and imprisonment of up to ten years.
The Federal Picture
Congress has yet to enact a comprehensive AI law, leaving the state patchwork as the operative reality for businesses. There is movement, however. In April 2026, a bipartisan group of four senators introduced the CHATBOT Act — the Children's Health, Advancement, Trust, Boundaries, and Oversight in Technology Act — one of the most detailed federal proposals aimed at how AI companies interact with children. It would mandate specific product features, account structures, parental rights, and data-use restrictions.
Yet the prospect of federal legislation that would preempt state rules remains uncertain. Until Congress acts, companies cannot rely on a single national standard to simplify compliance.
Why It Matters
For any company deploying a consumer-facing chatbot, the era of treating AI safety as optional is over. Obligations now apply immediately and vary by state, creating a fragmented compliance environment that depends on where a user is located rather than where a business is headquartered. A chatbot lawful in one state may violate disclosure or crisis-response requirements in another.
The deeper significance is what this wave reveals about how AI governance is actually taking shape in the United States. In the absence of federal action, states have become the primary regulators, and they are moving fastest where the harms are most tangible and the politics least divided — children's safety. That model, narrow and bipartisan, may prove a template for how other AI risks eventually get addressed.
For developers, the practical takeaway is to build disclosure, age-appropriate safeguards, and self-harm detection into products by default, and to treat the strictest applicable state standard as the baseline. The regulatory floor for conversational AI has risen sharply in 2026 — and every indication is that it will keep rising.