California AB 56 Explained: The Social Media Warning Law for Minors (Effective January 1, 2027)

California AB 56, the Social Media Warning Law, was signed by Governor Newsom on October 13, 2025 and takes effect January 1, 2027. It requires covered social media platforms — those that provide an addictive feed as a significant part of their service — to display a Surgeon General-style mental health warning to users under 18, modeled on tobacco warning labels, with screen-coverage and timing rules that escalate the longer a minor uses the platform in a single day. This guide explains exactly what AB 56 requires, how it interacts with the youth social media regime that already exists in California, why it matters for AI companion chatbots even though it is not nominally an AI law, and what platforms need to build in the eight-month window before the effective date.

What AB 56 actually does (and what it doesn't)

The most useful starting point is to clear up a common misconception. AB 56 is sometimes described in popular coverage as a "black box warning" for AI-driven content feeds, which is colorful but inaccurate. The phrase "black box warning" in AB 56 refers to the visual format of the notice — a boxed, prominent warning displayed to the user, deliberately modeled on the format the FDA requires for the most serious tobacco and prescription-drug warnings. It is not a reference to algorithmic black-box transparency. The warning AB 56 requires is about the mental health risks of social media use, particularly for children and adolescents, and it must be displayed using exact statutory text. Platforms cannot reword it, soften it, or pair it with reassuring counter-messaging on the same screen.

The statute itself is short and prescriptive, which makes it unusually clear by California-AI-law standards. Three things are required. First, the exact warning text — quoted verbatim in the FAQ below — must be displayed to users the platform reasonably determines to be under 18. Second, the warning must appear at three escalating triggers within a single day of use: on first daily access, after three cumulative hours of active use, and once per hour thereafter. Third, the warning must occupy specific minimum screen coverage and remain on screen for specific minimum durations, with the second and third warnings being non-dismissible. The non-dismissibility of the later warnings is the part most likely to provoke product-design pushback, because it actively interrupts use rather than just informing.

The mechanics: triggers, screen coverage, and timing

To understand why AB 56 will require real engineering work and not just a marketing copy update, it helps to walk through the three escalating warning states the statute defines. On a minor user's first access to the platform on any given calendar day, the platform must display the warning covering at least 25 percent of the screen for a minimum of 10 seconds. This first warning is dismissible — the user can click a clearly visible close affordance to continue. After that user has accumulated three hours of cumulative active use across whatever sessions they have on that day, the platform must display a second warning covering at least 75 percent of the screen for a minimum of 30 seconds. This second warning is non-dismissible. After every additional hour of cumulative active use beyond the initial three, the platform must display another 75-percent, non-dismissible warning. There is no daily ceiling on these later warnings — a user who spent six hours on the platform that day would see the warning four times: at first launch, at the three-hour mark, at the four-hour mark, and at the five-hour mark.

Three implementation details matter for platforms now. The phrase "cumulative active use" is not defined in the statute, which means platforms have to make a defensible engineering choice about what counts as active use — foreground time only, or all session time? Active scrolling, or also passive viewing? Background tabs? The statute's legislative findings emphasize features that are addictive by design, which suggests the legislative intent leans toward all foreground time. Until the Department of Public Health issues regulations or the Attorney General issues guidance, the safest reading is to count any time the user has the app or website open in an active foreground state.

The phrase "reasonably determines to be under 18" is also not defined, which is the central age-assurance question. AB 56 does not import a specific age-verification mechanism. In practice, platforms will use the same age-determination process they have been building for SB 976 (the Protecting Our Kids from Social Media Addiction Act, also signed by Newsom). That process can include a combination of self-declared age, behavioral signals, and parental account linking. Defensibility will turn on whether the platform's age-assurance method is reasonable given its user base — a platform with predominantly adult users gets more latitude than one with a teen-skewed audience.

How AB 56 fits into California's broader youth digital safety regime

AB 56 does not exist in isolation. It is part of a deliberate, multi-bill package that California has been assembling since 2022, and understanding the cluster makes the individual law much easier to comply with. The foundation is the California Age-Appropriate Design Code Act (AB 2273, signed 2022), which imposes affirmative duties on businesses that provide online services likely to be accessed by children — including data minimization, risk assessments, and high-default privacy protections. Layered on top is SB 976, the Protecting Our Kids from Social Media Addiction Act (signed 2024), which prohibits covered platforms from providing addictive feeds to minor users without verified parental consent and from sending them notifications during sleep and school hours. AB 56 is the third piece — the warning-label layer that informs minor users about the risks they face from a system that is already partially restricted by SB 976.

Read together, the regime works in a layered way. SB 976 restricts what platforms can do to minors. AB 56 requires platforms to warn minors about what they are doing. The Age-Appropriate Design Code requires that the underlying product itself be designed with minor safety in mind. Each law alone is incomplete; together they form a coherent framework that places California at the leading edge of youth digital safety regulation in the United States, alongside Minnesota (which enacted similar warning requirements) and a small handful of other states pursuing similar packages.

Why this matters for AI compliance even though AB 56 is not an AI law

A reader might reasonably ask why AB 56 belongs on a California AI compliance site at all. The honest answer is that AB 56 is not nominally an AI law, but its enactment carries direct implications for AI companion products and provides the regulatory blueprint for the AI-specific youth protection laws that are coming next. Three connection points matter.

First, the parallel bill. AB 1064 — the Leading Ethical AI Development for Kids Act — was the AI companion chatbot bill that was supposed to do for AI what AB 56 does for social media. AB 1064 would have imposed safety duties on AI companion apps used by minors, including content moderation requirements, suicide-risk safeguards, and disclosure obligations. Governor Newsom vetoed AB 1064 in the same legislative cycle that he signed AB 56, citing concerns about scope and feasibility. Bauer-Kahan and Bonta have publicly committed to bringing AI-companion legislation back, and they have explicitly named AB 56's framework as the model. Compliance teams at AI companion companies should read AB 56 not as a law that applies to them but as a near-future preview of a law that probably will.

Second, the definitional bleed. AB 56 borrows its "covered platform" definition from SB 976, which centers on the concept of an "addictive feed." That phrase was drafted with social media in mind, but the underlying mechanic — algorithmic content selection optimized for engagement — describes many AI companion products, recommendation systems, and conversational agents that operate at the boundary of social media. As enforcement actions surface and definitional disputes arise, expect product categories that did not consider themselves social media to be drawn into the orbit of these laws. AI companion apps with social discovery features, conversational agents that maintain ongoing relationships with users, and recommendation-driven AI products all sit in the gray zone.

Third, the design-pattern signal. The mechanics AB 56 prescribes — exact warning text, escalating screen coverage, time-based interruption, AG enforcement, no private right of action — are unusually specific compared with the broader transparency frameworks in SB 942, AB 2013, or SB 53. They look more like product regulation than like disclosure regulation. That is the design pattern AI companion regulation is likely to follow when it returns. Building products today that can flexibly accommodate prescribed text, time-based interruptions, and age-assured display logic is the right defensive posture even for companies not currently within AB 56's scope.

What covered platforms need to build before January 1, 2027

For platforms that fall squarely within AB 56's "covered platform" definition, the engineering work has four layers. The first is age determination — a reasonably reliable mechanism for identifying which users are under 18. Most large platforms already have this from SB 976 compliance work. The second is cumulative use tracking — the ability to count active use time per user per day, persisting across sessions and devices for logged-in users. The third is the warning UI itself: a 25-percent dismissible overlay for first-of-day, a 75-percent non-dismissible overlay at the three-hour and subsequent-hour triggers, with the exact prescribed text. The fourth is logging and audit — being able to demonstrate to a regulator that the warnings actually displayed to the right users at the right times, which means instrumenting display events the same way you would instrument any compliance-critical UI surface.

The non-engineering layer is policy. A written youth-safety policy that documents the platform's AB 56 implementation — covering age-determination methodology, cumulative-use definition, warning display infrastructure, and audit-logging — is the artifact regulators and enterprise partners will ask for in any compliance review. It is also what an Attorney General enforcement investigation would request first. Build the policy in parallel with the engineering work, not after.

What to watch over the next eight months

Three developments deserve close monitoring through the run-up to January 1, 2027. First, expect a First Amendment compelled-speech challenge to AB 56, likely filed by a major platform or a tech-industry trade association before mid-2026. The legal question will be whether the prescribed warning falls within Zauderer-style commercial speech doctrine (which permits compelled disclosure of factual, uncontroversial information) or triggers the heightened scrutiny that applies to compelled non-commercial speech. The case's outcome will materially affect whether the law takes effect on schedule. Second, watch for Department of Public Health regulations modifying the warning text — the statute authorizes such regulations, and changes between now and January 2027 are possible. Third, watch for the return of AI companion legislation. Bauer-Kahan has explicitly committed to bringing a successor to AB 1064, and that bill will likely incorporate AB 56-style mechanics. AI companion companies should treat the next legislative session as the moment when the regulatory shadow becomes substantive law.

Sources

The primary statute is AB 56 on California Legislative Information, with the Attorney General's position laid out in the Bonta press release on signing. Practitioner-grade analysis is available from Hunton Andrews Kurth and Mayer Brown. The legislative-history context, including the parallel AB 621 (deepfake nudification) bill that was signed alongside AB 56 and the AB 1064 AI companion bill that was vetoed in the same cycle, is documented in the Bauer-Kahan office press release. For ongoing tracking of any constitutional challenges and DPH regulatory activity, the California Attorney General's office is the most timely public source.