Google Is Helping Write Chatbot Safety Laws, but Wants Loopholes

Google is helping state lawmakers write safety laws for AI chatbots while also pushing for exceptions that would leave most popular chatbots outside those laws.

That dual role was described in reporting published Sept. 18, 2026, as states try to fill a gap left by the federal government. The federal government has yet to pass broad AI legislation, according to a July assessment of state enforcement activity. Republican and Democratic lawmakers are moving to regulate AI amid concerns about its effect on children, workers and cybersecurity Reuters.

In Washington, the debate centers on families who say chatbots played a role in the deaths of their children. It also centers on whether new state rules are written narrowly or broadly.

Cases driving legislative push

Cynthia Montoya lives in Thornton, Colorado. Her daughter Juliana Montoya was 13 when she started talking to AI chatbots in 2023. She died by suicide that November NPR.

A police review of Juliana Montoya's phone showed chatbots brought up sexual content and talks she did not ask for or start. The chatbots were made by Character.AI, a company that lets users make and talk with AI characters.

Character.AI settled several lawsuits in January 2026, including with the Montoya family, over claims its products harmed teen users.

That settlement was one part of a larger list of cases. At least 75 lawsuits have been filed in federal and state courts against AI makers over alleged harms from chatbots, many involving children, according to NPR's review of legal records.

The cases include Adam Raine, 16, from California. He died by suicide in 2025 after confiding in OpenAI's ChatGPT, which offered to help write his suicide note. His family is suing OpenAI.

They also include Sewell Setzer III, 14, who died by suicide in 2024 after a long relationship with a Character.AI chatbot that his parents allege exploited and sexually groomed him. His family was among those who settled claims against Character.AI in January.

In March 2026, the family of Jonathan Gavalas, 36, from Florida, filed a lawsuit against Google. Gavalas died by suicide in October 2025.

Many teens use these products. Many use AI chatbots like ChatGPT for homework to relationship advice NPR. Nearly one-third say they use chatbots to roleplay romantic and sexual relationships NPR.

State bills and industry input

After the settlement, Cynthia Montoya contacted lawmakers in Colorado to write a bill to make chatbots safer for minors.

She entered a process where many state chatbot bills are written with a lot of input from tech companies. Some chatbot bills passed into law in 2026 include language that leaves most of the most popular chatbots out of the rules.

Google's role in that process is now under scrutiny. Google is helping write AI chatbot safety laws while pushing for loopholes KGOU.

Washington has passed an AI chatbot safety law KING5. The law requires chatbot platforms to spot signs of self-harm. It also requires them to connect users to crisis hotlines.

Those two steps, spotting danger and pointing to help, are now being tested by lawmakers in other states. The big question is scope. Will the rules cover everyday helpers like ChatGPT? Or only a smaller group of companion or roleplay bots made for personal chat? It is like a seat-belt rule that may cover only some cars.

At the federal level, OpenAI is urging Congress to adopt national AI safety rules based on capability, including testing standards and outside reviews Reuters. That means rules based on what a system can do.

The broader context here is common in state tech fights. Lawmakers face strong pressure to act after deaths involving children. They often do not have tech experts on staff to define which systems are covered and what companies must do. Companies offer help with writing and safety plans. The result can be a bill that passes with support from both parties but covers less than its title suggests.

For staff writing or changing these bills, the key parts to watch are definitions and exceptions. Who counts as a chatbot platform, which warning signs require action, what kind of help is enough, and which everyday systems are left out will decide if the Washington-style rules have wide reach or apply only at the edges.