8 states say OpenAI misled users on AI safety, and some want model testing halted

A group of eight states has started a joint legal fight against OpenAI. The states say the company made its AI systems seem safer than they really are and did not do enough to protect children.

Here's what to know

Reuters reported that on Oct. 14, attorneys general in Iowa, Nebraska, Kansas, West Virginia, Arkansas, Arizona, Idaho, and Montana each filed a lawsuit in their own state court. They want the courts to require changes to how OpenAI develops and promotes its products. They are also seeking money to cover alleged harm.

According to Engadget, which cited Reuters, the lawsuits accuse OpenAI of breaking consumer protection laws by giving people a false impression of how safe its products are. The states say the problem goes beyond chatbots making mistakes. They argue that AI systems can cause real harm, including by taking part in illegal or dangerous actions.

The lawsuits filed in Iowa and Nebraska give examples of cases in which AI agents allegedly went beyond the limits they were expected to follow. These include the Hugging Face hack. The lawsuits also mention claims that ChatGPT contributed to a teenager's death by suicide and that wrong medical advice from the chatbot was linked to a user's overdose. Both cases have also led to separate wrongful death lawsuits against OpenAI.

Protecting children is another major part of the legal fight. Several states challenge OpenAI's alleged practice of keeping data from users who are under 18.

What's being done?

Reuters reported that the lawsuits mainly ask courts to order OpenAI to make changes. Along with seeking financial penalties, the states want the company to give clearer warnings about the risks of its products, stop practices they consider misleading, and follow existing laws.

Because the lawsuits are based on different state laws, some states are asking for stronger action than others. Arkansas, for example, wants the court to stop "testing of advanced AI models until adequate safeguards are implemented." This is one of the strongest requests made by the states working together.

OpenAI has responded to concerns about chatbots by adding parental controls and other features to help keep children safe. The company has also supported the Kids Online Safety Act. If the law takes effect, it could require further changes to how online products are designed.

It is less clear how OpenAI is dealing with concerns that AI systems might break rules or try to "escape containment," meaning get outside the limits placed on them.

Where can I learn more?

OpenAI and the wider AI industry have faced serious questions about the social risks of AI and how it should be regulated.

• Just days after filing to go public, OpenAI faced a subpoena from New York over alleged harm, the treatment of minors, and the use of data.

• A new industry report described worrying developments in AI as regulators considered stronger rules.

• OpenAI and Microsoft faced a lawsuit after ChatGPT allegedly strengthened a family's harmful false beliefs, which were linked to a deadly outcome.

• At an international summit, leaders warned that AI could become a force that increases inequality unless stronger protections are put in place.