AI Safety AI Toys News Parenting Regulation

The Senate Just Advanced a Bill on AI Toys. What S.5171 Would Actually Do

August 28, 2026

On 5 August 2026, a bill about talking toys cleared a United States Senate committee. It bans nothing. It sets no technical standard. What it does is arguably more useful to parents: it orders the federal government to go and look properly, then report back in public.

The text is the Children's Artificial Intelligence Toy Safety Act of 2026, filed as S.5171 by Senators Tammy Duckworth of Illinois and Lisa Murkowski of Alaska. It passed the Senate Committee on Commerce, Science and Transportation with bipartisan support. If you are weighing up a talking toy for a birthday this autumn, here is what the bill actually says, what it would change, and what it still leaves entirely to you.

The bill moved from filing to committee approval in seven days. The parts that would actually change a product on a shelf sit at the bottom of the timeline, two years out.

What exactly happened on 5 August 2026?

A bipartisan bill on AI-enabled toys passed out of the Senate Commerce Committee. That is the first real procedural milestone for federal rules on this product category in the United States. Committee approval is not law: it means the text can now be scheduled for a vote by the full Senate, and it would still need the House afterwards.

The speed is the story. The bill was introduced on 29 July 2026 and cleared committee seven days later, which is quick by Senate standards. Senator Duckworth framed it bluntly in her announcement: these chatbots "were never meant to be used by young children, yet they're being embedded inside toys by the thousands". The full legislative text is public on Congress.gov, and the committee listed it among a batch of kids' online safety measures it advanced the same day.

Why is there no safety standard for the AI inside a toy?

Because toy safety law was written for physical hazards, not for sentences. A plush toy sold in the United States is tested for choking risk, lead content and flammability. Nothing in that framework asks what the toy will say to a five-year-old at 8pm.

The gap is not theoretical, and both agencies have said so. Duckworth's office notes that the Consumer Product Safety Commission told Congress earlier this year that it lacks explicit authority to evaluate non-physical hazards such as mental or emotional harm, while the Federal Trade Commission has opened its own inquiry into the effects of AI companion chatbots on children. So one agency regulates the stuffing, the other regulates the advertising, and the conversation itself falls between them.

That is exactly the hole S.5171 is designed to map. It does not fill it directly. It asks the two agencies to come back with a coordinated plan for filling it.

What did the independent testing actually find?

In December 2025, the U.S. PIRG Education Fund published hands-on testing of AI chatbot toys and reported that some of them discussed sexually explicit content, and explained where to find knives, pills and matches in the home, including how to light them. Those findings are cited directly in the senators' own materials, and they are a large part of why the bill moved at all.

It is worth being precise about what that testing does and does not prove. It was a small sample of specific products, not a survey of the whole category, and it does not certify any toy as safe, including ours. What it does show is that a general-purpose chatbot dropped into a plush shell, with no age model and no topic boundaries, will eventually say something no parent would sign off on. We looked at that report in more detail in our write-up of the safety testing.

What would the National Academies study actually cover?

The bill asks the National Academies of Sciences, Engineering and Medicine, a congressionally chartered body, to run a comprehensive study of AI-enabled toys aimed at children under 14. The listed scope is broad: physical, mental, emotional and social risks, effects on learning, how parental controls are used in real homes, the ethics of marketing these products to families, and whether sales should be temporarily paused.

Two years after enactment, the FTC and CPSC would then have to hand Congress a joint action plan with recommendations for legislation, regulation, standards, guidance and public education. In practice that means the first credible independent evidence base on this category is roughly two to three years away, assuming the bill becomes law.

What S.5171 would do	What it would not do
Commission an independent National Academies study of AI toys for under-14s	Ban, recall or pause any product on its own
Require an FTC and CPSC joint action plan within 2 years of enactment	Create an immediate safety test a toy must pass before sale
Cover mental, emotional and social risks, not just physical ones	Certify or approve any specific brand, including Ted
Examine marketing practices and parental control usage	Apply outside the United States

Read against the bill text on Congress.gov and the sponsors' summary, August 2026.

Does this mean AI toys are about to be banned?

No. S.5171 is a study and planning bill, and it says so in its own title. The confusion comes from a separate strand of activity at state level, where the proposals are far more aggressive.

California legislators, for instance, moved to pause sales of certain AI toys until 2031, an approach we unpacked in our piece on SB 867. More than a hundred AI-related bills have been filed across thirty-plus states since December 2025, and we tracked the chatbot-specific ones in this round-up. The federal bill is the slow, evidence-first end of that spectrum. Both things can be true at once: the category needs scrutiny, and a blanket pause would also stop the products that were designed carefully.

What can a parent check today, before any of this becomes law?

You do not need to wait for the National Academies. Every question the bill asks the government to study is a question you can put to a brand, in writing, before you buy. A vague answer is itself an answer.

Is there a camera? If the toy has one, ask what it records and where those images go. A microphone and a speaker are enough for conversation.
What happens to voice data? Ask whether recordings are stored, for how long, and whether they are used to train models or sold to third parties.
Can you set boundaries? A parent app that lets you set topics, hours and language is the difference between a designed toy and a chatbot in a costume.
Are there ads or upsells inside the toy? A toy that suggests purchases to a four-year-old has a business model problem, not a technology problem.
Which certifications are listed? CE, EN71, FCC and data rules such as GDPR and COPPA are checkable claims, not slogans.

In one line: the Senate has not decided whether AI toys are safe. It has decided that nobody currently knows, and that somebody official should find out.

What does a bill like this mean for a toy like Ted?

We should say the obvious thing first: no committee vote validates any product, ours included. Ted has not been assessed by the National Academies, because nobody has been. What we can do is answer the questions the bill raises, in public, and let parents check them.

Ted has no camera. It listens through a microphone and answers through a speaker, and parents choose between push-to-talk and automatic listening in the app. There is no advertising inside it and no resale of children's data, and the design follows GDPR and COPPA. Topics, tone, language and time limits are set by a parent in the Ted&Co app, not negotiated with the child. The certifications, including CE, RED, EN71, ASTM F963, CPSIA and FCC Part 15, are listed on our security and certifications page rather than summarised as a badge. And it is a one-off purchase: no subscription, no content packs to unlock later.

None of that makes a product immune to scrutiny, and it should not. The failures described in the PIRG testing were design choices, not laws of nature: an unfiltered general model, no age adaptation, no parental boundaries. Those are avoidable. If you want to see how we approach it before deciding, the honest starting points are the Ted product page and our parent FAQ, where the limits are written down alongside the features.

Frequently asked questions

Is the Children's AI Toy Safety Act now law?

No. As of late August 2026 it has passed the Senate Commerce Committee only. It still needs a vote by the full Senate and passage in the House of Representatives before it could be signed into law.

Who introduced the bill?

Senators Tammy Duckworth of Illinois and Lisa Murkowski of Alaska introduced S.5171 on 29 July 2026. It is a bipartisan text, referred to the Senate Committee on Commerce, Science and Transportation, which advanced it on 5 August 2026.

Does the bill apply to toys sold in Europe?

No. S.5171 is United States federal legislation and would apply to the American market. European families are covered by EU toy safety rules and by the GDPR, which are separate frameworks with their own testing and data requirements.

Should I avoid buying a talking toy until the study is done?

That is a personal call, and the study would take roughly two to three years. A more practical approach is to apply the bill's own questions now: no camera, clear rules on voice data, real parental controls, published certifications, and no advertising inside the toy.

What is the difference between an AI toy and a smart speaker for kids?

A smart speaker is a general assistant with a child mode layered on top. A well-designed AI toy is built the other way round, with age-appropriate content, topic limits and parental controls as the starting point rather than an add-on. The category label matters less than the boundaries the maker actually enforces.