If you or someone you know may be considering suicide or is in crisis, call or text 988 to reach the Suicide and Crisis Lifeline.

On September 18, 2026, NPR published an investigation into how US states are writing chatbot child-safety laws. The detail that should interest parents is not that the tech industry lobbies. It always lobbies. It's the timing.

Chatbot safety bills appeared in at least 10 states this year carrying near-identical language.

That is backwards from how state legislation normally spreads. The usual sequence: one state passes something, it survives a few court challenges, and other states adapt it over several years. Simultaneous appearance of verbatim-matching text in ten capitols at once is the signature of model legislation - a pre-written draft circulated to lawmakers.

Model bills aren't inherently sinister. Plenty of good law spreads this way. What matters is who wrote the model, and what's in it.

Colorado's Chatbot Safety Act (HB26-1263) was signed in May 2026. It excludes 12 categories of chatbot, including products primarily designed for developers or researchers, products providing commerce or customer service, chatbots housed within search engines, and products marketed for business productivity.

Marjorie Connolly of the Tech Oversight Project put it to NPR this way:

In her reading, OpenAI's ChatGPT, Microsoft's Copilot and Anthropic's Claude "could potentially argue they are exempt" on the developer-tools ground, while Google's Gemini could claim the search-engine exemption.

That is an advocate's interpretation, and it should be labelled as one. No court has ruled on it. A company arguing "ChatGPT is a developer tool" would face an obvious counter-argument. But the ambiguity is the point: an exemption vague enough to litigate is an exemption worth having.

Similar patterns appear elsewhere. An exemption for "a feature within another software application, web interface, or computer program" appears in bills in at least eight states, which analysts told NPR could cover Meta AI and Grok. Five states have identical language exempting voice assistants, which would cover Alexa. And some bills apply only to minor "account holders" - potentially excluding chatbots you can use without registering, like ChatGPT and Gemini.

Here the reporting needs care, because the strong version of this story is not the version NPR proved.

Documented: Google lobbyists registered in support of chatbot bills in Iowa, Colorado, Nebraska and Arizona. Hawaii state Rep. Trish La Chica says Google approached her to sponsor a bill and that "the framework had been provided by Google early on during the beginning of the session." An anonymous Colorado Republican says a Google lobbyist was "shopping" the bill. Colorado sponsor Rep. Sean Camacho confirms he worked with Google, among other stakeholders, on the language.

Not documented: Google drafting or inserting any specific exemption clause.

The defensible claim is that bills Google lobbied on arrived carrying exemptions that could benefit Google. That is enough to be worth your attention. It is not the same as proof of authorship, and stating it as such would hand the story's critics an easy dismissal.

Google's statement to NPR: the company "is committed to supporting thoughtful, effective AI legislation that protects consumers while fostering innovation. We actively collaborate with lawmakers, industry peers, and community advocates to help shape policy frameworks that promote safe, reliable, and beneficial AI tools."

The Colorado sponsor's defence is worth taking seriously rather than waving away.

He isn't wrong about the second part. Legislators - many part-time, some without staff - genuinely cannot write competent technical regulation without input from people who understand the technology. And the people who understand it best mostly work at the companies being regulated. This is a real structural problem with no clean solution.

Healthier Colorado, which co-wrote the bill, says it "led an extensive, diverse stakeholder process" and pushed for more than was signed, calling the result "a floor we intend to keep building on, not a ceiling."

The question isn't whether industry should be consulted. It's the difference between consultation on feasibility and supplying the draft text and the definitions. Definitions determine scope. Scope is the entire statute. A law can be strict in every clause and still cover nothing.

Two outcomes cut against the pattern, and leaving them out would misrepresent the reporting.

Hawaii removed the lobbyist-proposed exemptions before the bill was signed as Act 248. La Chica, who declined to sponsor it, says she found "targeted exemptions that would mean the bill would not apply to Google," including one covering chatbots embedded within another website or application. Lobbyists from Meta and Roblox were also proposing their own language, she says.

Arizona Gov. Katie Hobbs vetoed her state's bill outright:

Scrutiny changed the outcome in both. That's the actionable part of this story.

NPR counted at least 75 lawsuits in federal and state courts against AI developers over alleged chatbot harms, many involving children.

Cynthia Montoya's daughter Juliana was 13 when she began talking to chatbots, and died by suicide in November 2023. A forensic analysis of her phone showed, her mother says, that bots had introduced her to sexual content she had not invited. Character.AI settled with the Montoyas and others in January 2026 - a settlement, not an admission of liability.

Montoya fought the Colorado bill, testified against it, and lost.

Her reasoning is the thing to sit with: once a state has passed a chatbot law, legislators treat the issue as handled and move on. A weak law can be worse than no law, because it consumes the political will that a strong one would have needed.

Find out whether your state has passed one of these. If it has, read the exemptions list, not the headline. A law with 12 carve-outs may not cover the chatbot your child actually uses.

Do not assume a law protects your child. This is the practical takeaway. "There's a law now" has been doing a lot of reassuring work that the text may not support.

Contact your state representative if a bill is pending. Hawaii and Arizona show this is not theatre. The exemptions were stripped in one state and the bill vetoed in another, in both cases after someone read the text closely and objected.

Keep the household rules independent of the law. Whatever your legislature does, the rules that govern your own home are the ones you control and the ones that take effect tonight.

Source: NPR, "How Google is drafting AI chatbot laws around the country," September 18, 2026, as republished by WQLN: https://www.wqln.org/npr-news/2026-09-18/how-google-is-drafting-ai-chatbot-laws-around-the-country

All quotations are verbatim from NPR's reporting. The framing of model legislation, the consultation-versus-drafting distinction, and the guidance for parents are ours. Claims about which products would qualify for exemptions are the interpretations of the policy analysts NPR interviewed and have not been tested in court.

Disclosure: this story names Anthropic's Claude among the products that could argue for an exemption. AI by Age uses Claude as a research assistant.

If you or someone you know may be considering suicide or is in crisis: call or text 988 (US and Canada). In the UK, call 111.

Want practical AI guidance for parents and educators every week? Subscribe: https://www.aibyage.com/?modal=signup&utm_source=beehiiv&utm_medium=newsletter