In 1996 Congress decided that the place to filter children's television was the television set. Section 551 of the Telecommunications Act led the Federal Communications Commission to require a V-chip in every set with a picture 13 inches or larger made from 1 January 2000. The broadcasters encoded ratings, the manufacturers built the decoder, and parents held the switch. Thirty years later Florida's Attorney General has sketched a similar division of labor for the phone, with one difference that matters: the device is also the place where a child talks to an AI system.
On 22 October 2026 Attorney General James Uthmeier released the Safe Screens Act proposal, a one-page summary of legislation aimed at app stores, social media platforms and AI chatbots. The release quotes Polk County Sheriff Grady Judd and Polk State College President Anastasios Camoutsas. There is no bill number yet, and no published statutory text. What exists is a list of duties, and the list is specific enough to read closely.
Practical takeaway. The Safe Screens Act is a proposal, not a filed bill. As released on 22 October 2026, it would require app stores and AI chatbot operators to determine a user's age, make operating systems apply device-level restrictions by age, bar a defined set of chatbot features for all minors, and enforce the whole package through Florida's existing social media statute at up to $50,000 per violation.
Which companies the 22 October 2026 release would make determine a user's age
The release names three layers. App stores "and similar marketplaces must determine the user's age, link minors to a parent's account, and obtain parental consent before a minor downloads or buys an app from a covered platform." AI chatbot operators and platforms "must determine account holder's age." Operating systems "must apply age-appropriate restrictions at the device level based on the user's age." A single sentence then covers all three: "All age determination must use commercially reasonable, privacy-protective methods."
Read literally, the operating system is assigned restrictions, and the duty to determine age sits with the app store and the chatbot operator. The release does not say where the operating system obtains the age it acts on. Tech Times, writing on 26 October 2026, read the clause as requiring every adult smartphone user in Florida to prove their age to the phone's operating system. That may be how a drafted bill ends up working, since the app store and the operating system are run by the same two companies on most phones. The release itself leaves the route open, and the drafting choice decides which company establishes a user's age, what data it needs to do so, and which other companies may rely on the result. Collecting a birthdate is one possible method; the release prescribes none.
California has already made that choice. Assembly Bill 1043, the Digital Age Assurance Act signed on 13 October 2025 and operative from 1 January 2027, in its text as enacted in 2025, has the operating system collect an age or birthdate at account setup and pass an age-bracket signal to developers who request it. Later California amendments are not reviewed here. Florida's release neither adopts nor rejects that model.
What the proposal would bar AI chatbots from doing with users under 18
The chatbot section applies to "all minors," a wider group than the 14-to-17 teen accounts addressed in the social media section. Operators could not provide minors features that "simulate companionship through personification, emotional engagement, user-data retention, unsupervised therapy, sexual content, or other deceptive features." They could not encourage self-harm, suicide, eating disorders or illegal activity, generate or facilitate sexually explicit conduct, "suggest the chatbot is a real human," "prioritize validation or engagement over user safety or guardrails," or encourage isolation from appropriate adults.
Two items in that list reach into product architecture. User-data retention appears as a companionship feature, which places a chatbot's memory of earlier conversations in the same category as a persona. A drafted bill would have to say whether that covers all saved history or only memory used to sustain a relationship. The phrase "prioritize validation or engagement" names a training and tuning property, the tendency of a model to agree with its user, and makes it a legal exposure when the user is a child. Both depend on the operator knowing which accounts belong to minors, which is why the age clause and the feature list arrive together.
The human-impersonation ban has a counterpart in Europe. Article 50(1) of the EU AI Act requires providers to design systems so that people are told they are interacting with an AI system unless that is obvious, a duty Quentir set out in its post on what Article 50 requires and when. Florida's version protects minors only and bans the claim of being human, with no duty of affirmative notice. The EU's Digital Services Act, in Article 28, separately asks online platforms accessible to minors for measures ensuring a high level of privacy, safety and security. A chatbot service offered in both markets could meet three overlapping rules with three different triggers, depending on its role under the AI Act and on whether it qualifies as an online platform under Article 3(i) of the DSA, since Article 19 exempts micro and small enterprises from Article 28 unless they are designated as very large online platforms.
How Florida's 2024 social media law, section 501.1736, would carry the enforcement
The proposal does not build a new enforcement regime. Knowing or reckless violations would be unfair practices under section 501.1736 of the Florida Statutes, the provision titled "Social media use for minors" that HB 3 created in 2024. That section already requires social media platforms to bar account holders younger than 14 and lets the Department of Legal Affairs collect a civil penalty of up to $50,000 per violation. The release extends those remedies "equally to covered app stores, social media platforms, and AI chatbot operators" and adds a public website on which the Attorney General's office would track incidents.
The timing follows the courts. Uthmeier's statement says "the Eleventh Circuit just upheld that law." Tech Times dates the ruling to 14 October 2026 and describes it as a finding that HB 3 is likely constitutional. A preliminary ruling of that kind supports the existing statute; it says nothing yet about extending the statute to operating systems and chatbots, which raise different questions about compelled design and the speech of adults who would pass through the same age checks.
The social media section is the most detailed part of the release. Teen accounts would get a quiet mode from 11 p.m. to 7 a.m. all year and from 8 a.m. to 3 p.m. on school weekdays, a limit of two hours a day and 60 hours a month including messaging, pauses at 60 and 90 minutes, hidden like and comment counts, autoplay and infinite scroll off by default, cosmetic-procedure filters disabled, and a prompt every three days toward a non-personalized feed that teens cannot dismiss. Reports of child sexual exploitation would need human review within 48 hours.
How Quentir Reads It
The V-chip rule named one decoder in one device. The Safe Screens Act names three parties and gives one of them, the operating system, a duty to act on information the release assigns to the other two. Every legislature that moves age assurance down to the device meets the same engineering question: the law has to say when receiving an age signal produced in one layer counts as knowledge for every layer that reads it, and when a recipient may rely on it. California, in AB 1043 as enacted in 2025, answered by making the operating system the source. Florida's summary has not answered yet, and its 2027 session will show whether a filed bill does.
For AI developers the more durable signal is the feature list. State attorneys general have so far acted against chatbot companies through lawsuits and petitions. This proposal turns specific product behaviors (persona, memory, agreement-seeking and the claim to be human) into defined prohibitions for minors, and attaches them to a statute with a per-violation penalty that a court has just allowed Florida to enforce. If enacted in this form, families would gain a clear rule on what a chatbot may not do with a child. Adults might pay for it with one more checkpoint between them and their own device, depending on how the privacy-protective standard is written.
The disclosure side of this problem already has a fixed text in Europe. Quentir's Signature Brief 2026.3, AI Act Article 50 Disclosure, maps that duty across provider and deployer roles with a dated source spine, an executive summary and an internal-use checklist; this post covers only the Florida proposal and where it touches Article 50.
Florida's 2027 regular session opens on 2 March 2027. The first test of the proposal is whether a filed bill keeps the operating-system clause, and whether it says which company establishes an adult's age, and by what method.
Sources: Office of the Attorney General of Florida, Attorney General James Uthmeier Proposes the Safe Screens Act to Protect Minors on App Stores, Social Media Platforms, and AI Chatbots, release dated 22 October 2026; The Florida Senate and Legislature, Florida Statutes section 501.1736, Social media use for minors (2026 Florida Statutes); California Legislature, AB 1043, Digital Age Assurance Act, chaptered October 2025; Congressional Research Service, V-Chip and TV Ratings: Helping Parents Supervise Their Children's Television Viewing (report 97-43, 2002 update); Regulation (EU) 2024/1689 (AI Act), Article 50(1); Regulation (EU) 2022/2065 (Digital Services Act), Articles 3(i), 19 and 28; Tech Times, Florida AG's Safe Screens Act Would Force All Smartphone Users to Verify Age at OS Level, 26 October 2026 (secondary coverage, compared with the primary release). No bill text or bill number had been published when the release was checked on 26 October 2026.
Published intelligence, built to inform your own decisions. Published: October 26, 2026.