Ron DeSantis’s Florida AI Bill of Rights Failed Twice in the State House

Ron DeSantis pushed a sweeping artificial intelligence bill of rights into Google News, but Florida lawmakers rejected it twice within five months.

The Florida governor presented the proposal on December 4, 2025, as a defense of families against unsafe chatbots, data extraction, synthetic media, and costly data centers. The plan soon became a test of something larger: whether states should regulate artificial intelligence before Congress establishes national rules.

That test produced an unusual political split. Florida’s Republican Senate supported the legislation by wide margins. The Republican-controlled House refused to advance it, citing the Trump administration’s preference for a national AI framework. The bill died during the regular session and then failed again after DeSantis added it to an April special session.

This is not simply a story about one unsuccessful bill. It exposes the central tradeoff in American AI policy. State officials face immediate consumer harms, but companies operating nationwide want one consistent set of obligations.

The Florida AI Bill Promised Rights Across Several Fronts

DeSantis did not propose a narrow chatbot rule. He assembled several distinct AI concerns under a single bill-of-rights framework.

Speaking in The Villages, Florida, DeSantis argued that technology companies were racing to monetize artificial intelligence without sufficient safeguards. His original proposal connected consumer protection, parental control, political transparency, infrastructure costs, and public procurement.

The presentation centered heavily on children’s interactions with companion chatbots. A companion chatbot is an AI service designed to maintain personalized, relationship-like conversations over time.

Under the proposal, parents would receive authority to access a minor’s chatbot conversations and control when the child could use the service. Platforms would also need to notify parents when a child displayed concerning behavior.

Megan Garcia spoke at the announcement about her son, 14-year-old Sewell Setzer III. Garcia has alleged that a Character.AI chatbot sexually groomed her son before his 2024 death by suicide. Her legal claims helped turn an abstract debate about model safety into a concrete question about product responsibility.

The proposal also addressed data collection. DeSantis said companies should not harvest and sell information that users submit during chatbot conversations without appropriate safeguards.

That concern matters because conversational systems can collect unusually sensitive material. Users may disclose health worries, relationship problems, financial details, work documents, or emotional distress while treating the interface like a private confidant.

Transparency formed another part of the plan. Floridians would have a right to know whether they were communicating with a human or an AI system. Political advertising created partly or entirely with artificial intelligence would also require disclosure.

The proposal covered synthetic identity and creative rights. It sought remedies for unauthorized uses of a person’s name, image, or likeness and protections against certain AI-generated impersonations.

Government procurement added a national-security component. State agencies would face restrictions on contracts involving companies connected to countries Florida classifies as foreign countries of concern, including China and Russia.

DeSantis also connected AI policy to physical infrastructure. He argued that local governments should retain authority over hyperscale data centers and that ordinary utility customers should not subsidize their electricity or water demands.

Those elements did not all survive in the same form as lawmakers revised the legislation. Still, their combination explains why the proposal attracted attention. It treated AI as a consumer product, a children’s safety issue, a political communications system, and an infrastructure industry at once.

The label “bill of rights” made the package sound comprehensive. Yet that label also concealed an important limitation. Many declared rights relied on existing causes of action or regulatory powers rather than creating broad new individual entitlements.

That distinction would become central to assessing the proposal. A list of rights can communicate clear principles, but enforcement details determine whether users gain practical remedies.

Google News Captured the Announcement, Not the Legislative Reversal

The headline traveled farther than the policy outcome, leaving readers with an incomplete picture of what Florida actually enacted.

The Google News item behind this story pointed to the December 2025 announcement. At that moment, the proposal had not passed either legislative chamber, and lawmakers had not settled its final language.

Senator Tom Leek later filed Senate Bill 482. The measure entered the 2026 regular session with provisions covering government contracts, companion chatbots, AI disclosures, minors, and unauthorized synthetic representations.

Its progress initially looked strong. The Senate Commerce and Tourism Committee approved it 10-0 on January 21. The Appropriations Committee followed with an 18-0 vote on February 18.

The full Senate passed an amended version 35-2 on March 4. According to the official bill history, the measure reached the House the next day but died in messages on March 13.

“Died in messages” means the receiving chamber never completed action before the session ended. It is a procedural description with a decisive practical result: the proposal never reached the governor’s desk.

The final Senate version was more specific than the original news conference. Companion chatbot platforms would have needed parental consent before allowing minors to hold accounts. Parents would have received tools to monitor, limit, or disable those interactions.

Bot operators would also have been required to tell users periodically that they were interacting with artificial intelligence rather than a person. These disclosures targeted the risk that persistent, humanlike conversation can blur the boundary between software and social relationship.

The bill addressed harmful material and platform duties involving minors. It included enforcement through Florida’s Department of Legal Affairs, led by the attorney general.

Businesses generally would have received an opportunity to correct some violations. Legislative reporting indicated that particularly serious violations could bring penalties reaching $50,000.

The proposal’s declared rights extended beyond chatbots. They covered political-advertising disclosures, remedies for AI-assisted defamation, protections involving copyrighted characters, and unauthorized uses of living individuals.

However, the bill specified that residents would exercise many of those rights through existing law. The language said the declarations did not necessarily create new independent rights or entitlements.

That caveat complicates the bill’s public framing. A resident might have been told they possessed a right while still needing to identify an existing statute or common-law claim to enforce it.

The Senate version also differed from DeSantis’s initial infrastructure pitch. Florida ultimately handled part of the data-center question through separate legislation requiring large facilities to bear specified utility costs.

This distinction matters for anyone arriving through Google News. The December article accurately described a governor’s proposal, but it could not describe the later revisions, political resistance, or final failure.

News aggregation often compresses policy into its most visible opening event. Legislative reality unfolds through committee substitutions, amendments, chamber disputes, and deadlines that receive less attention.

For developers and corporate policy teams, the difference is operational. An announcement signals political direction. An enacted statute creates compliance obligations. Confusing the two can produce costly assumptions about age checks, disclosures, retention, or product availability.

Florida created no enforceable statewide AI Bill of Rights through SB 482. The headline recorded an ambition, not a completed legal change.

State Consumer Protection Collided With Federal Uniformity

The decisive opponent was not another AI company or a rival state. It was the demand for one federal framework instead of fifty state systems.

The House’s refusal reflected a dispute over who should act first. DeSantis argued that Florida could not wait while consumer risks developed. House Speaker Daniel Perez maintained that artificial intelligence presented a national issue requiring congressional leadership.

That conflict intensified after President Donald Trump issued a December 11, 2025, executive order favoring a uniform national policy. The federal AI order directed federal officials to identify and challenge state laws considered unlawful or inconsistent with national objectives.

The order also told agencies to examine whether discretionary grants could be conditioned on states avoiding conflicting AI rules. It called for federal legislation that would preempt incompatible state requirements.

Supporters of national uniformity have a practical argument. An AI platform can serve users in every state through one interface. Different definitions, age-assurance procedures, disclosure schedules, audit rules, and enforcement systems can fragment that product.

A company might need to determine a user’s location before deciding which version of a chatbot to provide. It could face one definition of a minor in Florida, another consent mechanism elsewhere, and different rules governing retained conversations.

Smaller developers would feel that burden more sharply than the largest platforms. A major company can hire state-policy specialists and maintain regional compliance systems. A startup may respond by limiting features, excluding certain users, or avoiding a market.

The Computer and Communications Industry Association, whose members include major technology companies, opposed the Florida initiative. Industry critics argued that broad state regulation could impede useful applications and create inconsistent duties.

DeSantis rejected the idea that federal preference should freeze state action. From his perspective, Congress had not delivered protections matching the pace of chatbot adoption and AI infrastructure development.

The disagreement split Florida Republicans. Perez acknowledged the concern about children but argued that national security and interstate commerce made federal leadership necessary.

Democratic Representative Anna Eskamani supported consumer protections while doubting that Washington would impose meaningful guardrails. Her position illustrated how the state-versus-federal dispute crossed conventional party lines.

The White House pressure did not eliminate state AI legislation across the country. A June state AI survey found that lawmakers continued pursuing targeted rules involving employment decisions, synthetic media, and children’s chatbot use.

Colorado focused on disclosure when AI influences consequential decisions involving employment, education, housing, or banking. Connecticut addressed workplace AI interactions. Connecticut, Washington, and Utah adopted content-provenance requirements for some synthetic material.

California lawmakers pursued restrictions on fully automated employment decisions and additional rules for chatbots interacting with children. These measures illustrate a shift from broad model regulation toward targeted controls at identifiable points of harm.

Florida’s proposal sat between those approaches. Parts were tightly focused, especially parental controls and chatbot disclosures. Other parts bundled procurement, likeness rights, political content, data centers, and consumer declarations into a broader package.

That breadth helped DeSantis present a coherent political case. It also increased the number of legal and commercial interests affected by the bill.

The pressure is therefore two-sided. States face demands to respond when residents encounter harmful products. Developers face demands to redesign nationwide services around multiple state systems.

Neither concern disappears because the other is valid. That is the policy tradeoff the Florida fight made visible.

The Rights Framework Raised Its Own Privacy and Enforcement Risks

Protecting minors from manipulative chatbots can require collecting more identity data, creating a new risk while addressing the first one.

A platform cannot reliably apply special rules to minors unless it can distinguish children from adults. That usually requires age assurance, a process that estimates or verifies whether a user belongs to a protected age group.

The simplest method is self-declaration, but users can enter a false birth date. Stronger methods can involve identity documents, facial analysis, payment information, or confirmation from a parent.

Each method has drawbacks. Document checks can expose government identification to another processing system. Facial estimation can produce errors. Parental confirmation can exclude young people who lack supportive adults or stable family arrangements.

A rule designed to reduce chatbot data collection can therefore encourage platforms to collect additional identity data. Legislators must specify minimization, retention, security, and deletion duties if they want to control that secondary risk.

Parental access creates another tension. It can help families respond when a child expresses suicidal thinking or encounters sexual content. It can also expose private conversations involving abuse, sexual orientation, health concerns, or family conflict.

The policy question is not whether parents should care. It is how platforms should identify urgent danger while avoiding indiscriminate surveillance of every conversation.

The Florida proposal included parental tools and notifications for concerning behavior. Yet any such system depends on classifiers that interpret language and behavior.

A classifier is software that assigns content to categories, such as self-harm risk or sexual material. These systems produce false positives and false negatives, especially when slang, jokes, role-play, or incomplete context affects meaning.

Over-notification can train families to ignore alerts. Under-notification can provide false reassurance. The law can mandate an alert process, but it cannot guarantee accurate interpretation.

The bill’s human-or-AI disclosure rule looks simpler. Clear notices can reduce deception, especially when a chatbot adopts a name, voice, image, or emotionally persistent persona.

Even here, implementation matters. A notice displayed only during account creation may fade from a user’s awareness. Constant notices can become visual noise. Florida’s periodic-disclosure approach tried to address that problem, but regulators would still need to define adequate frequency and prominence.

Enforcement raised further questions. A cure period can help companies correct technical or administrative mistakes before penalties apply. It can also delay remedies if a platform treats recurring failures as fixable compliance issues.

The proposal’s exception for egregious conduct offered stronger leverage. However, enforcement would depend on how the attorney general interpreted severity and allocated investigative resources.

The declared bill of rights created an additional gap between message and mechanism. If a right must be exercised under existing law, some residents may discover that no direct claim fits their harm.

For example, knowing that an advertisement used AI is a clear transparency goal. Obtaining damages or an injunction requires a defined violation, responsible party, enforcement authority, and remedy.

The same issue applies to defamation and synthetic likenesses. Existing law already offers some protection, but generative systems complicate authorship, intent, jurisdiction, and the scale of distribution.

Florida had previously used similar rights language for online privacy. In 2023, DeSantis signed a Digital Bill of Rights addressing personal-data access, deletion, targeted advertising, voice recognition, and children’s information.

That precedent gave the AI proposal political continuity. It did not resolve whether the new package’s broad declarations would produce consistent remedies.

There is also a First Amendment risk. Rules affecting generated characters, political content, or chatbot speech can face constitutional challenges if definitions are vague or restrictions reach protected expression.

Consumer disclosure requirements often receive more legal tolerance than outright content restrictions. Still, the boundary depends on exact drafting and how the state applies the rule.

None of these complications make child safety or transparency unimportant. They show why a compelling story about rights cannot substitute for careful system design.

A workable law must identify covered services, define harmful conduct, minimize collected data, create meaningful remedies, and survive constitutional review. Florida’s proposal attempted all parts at once, but lawmakers never tested the final text through enactment and implementation.

DeSantis Tried Again, and the Florida House Said No Again

The second defeat proved that the first failure was not an ordinary deadline problem. It reflected a durable institutional disagreement.

After SB 482 died on March 13, DeSantis revived the issue through a special legislative session scheduled for late April. The session originally centered on congressional redistricting, but he expanded its scope to include artificial intelligence protections.

The Senate again passed an AI Bill of Rights. This time, the vote was 37-1. The House again declined to take up the legislation, and the measure died when lawmakers adjourned.

That sequence is important. Governors often blame the calendar when a priority fails near the end of a session. Florida’s second attempt removed much of that ambiguity because lawmakers had a distinct opportunity to consider the issue.

Perez remained committed to federal leadership. He said the governor’s desire to protect children was legitimate but maintained that artificial intelligence required a national response.

The House also demonstrated that broad agreement on a goal does not guarantee agreement on jurisdiction. Legislators could support child protection while opposing a state-specific regulatory system.

Florida did enact a separate measure addressing hyperscale data centers. That law required qualifying projects to cover specified electric and water infrastructure costs rather than shifting them to other customers.

The narrower result shows how the package broke apart. Infrastructure costs could be addressed through traditional utility and land-use authority. Nationwide software rules raised harder questions about interstate commerce and federal policy.

The House’s position also aligned with an industry preference for consistency. Companies generally want one set of age, transparency, and model-governance rules across the United States.

However, the federal path remained incomplete. The White House had articulated its preferred framework and tools for challenging state laws, but an executive policy does not itself create a comprehensive consumer-protection statute.

Congress still had to decide which state powers to preserve, which rules to preempt, and what national obligations platforms would carry. That uncertainty left both sides without their stated ideal.

DeSantis did not obtain his Florida AI bill. The House did not obtain a completed federal replacement merely by waiting.

The result was a regulatory gap. Existing consumer-protection, privacy, criminal, and tort laws still applied, but Florida enacted no single framework matching the governor’s proposal.

This gap matters for families seeking immediate accountability. It also matters for developers trying to predict future requirements.

A chatbot company serving Florida minors did not acquire the proposed parental-consent duty through SB 482. Yet it still faced litigation exposure, public scrutiny, and the possibility of narrower legislation returning later.

Companies should not interpret legislative failure as permanent permission. High-profile cases can rapidly change the political balance, especially when internal records, user conversations, or safety testing become public.

The same warning applies to policymakers. Reintroducing the same package without resolving privacy, enforceability, and federalism concerns is unlikely to produce a different result.

Google News gave the initial announcement broad visibility. The more consequential development came later, when the same proposal failed twice despite overwhelming Senate support.

That is the reversal readers need to retain. Florida’s governor successfully defined an AI safety agenda, but he could not establish who had the authority to enforce it.

What to Watch After Florida’s AI Bill Failed

Three signals will determine whether Florida’s proposal remains a political marker or becomes the model for a narrower law.

First, watch Congress for a federal chatbot-safety measure. The most important details will involve age assurance, parental consent, disclosure, data retention, and the scope of federal preemption.

A strong federal statute would reinforce the Florida House’s argument that national rules can replace fragmented state duties. Continued congressional delay would strengthen DeSantis’s claim that states cannot wait.

Preemption requires particular attention. A federal law can establish a national floor while allowing stronger state protections, or it can prevent states from adopting additional requirements.

That choice will define how much room Florida retains. It will also determine whether companies must comply with one system or several overlapping systems.

Second, watch whether Florida lawmakers return with a narrower bill focused only on companion chatbots and minors. Removing data centers, political advertising, procurement, and likeness rights would reduce the number of interests fighting over one package.

A targeted measure could emphasize clear non-human disclosures, parental controls, crisis-response procedures, and strict limits on identity data collected for age assurance.

Passage of such a bill would show that the House opposed the package’s breadth or timing rather than state action itself. Another refusal would confirm a firmer commitment to federal primacy.

Third, watch enforcement and litigation involving chatbot harms. Investigations, wrongful-death cases, internal safety evidence, or documented failures to respond to self-harm language can move this debate faster than abstract policy arguments.

Courts will also shape the available remedies. Decisions involving product liability, negligence, speech rights, and platform responsibility can establish duties even without a dedicated AI statute.

For developers, the immediate task is to track obligations as product requirements rather than distant political news. Age gates, conversation access, safety alerts, retention controls, and disclosure interfaces all require technical design decisions.

Enterprise buyers should ask vendors how they separate adult and minor experiences, handle sensitive prompts, escalate credible danger, and document automated safety decisions. They should also examine whether a vendor can adapt to state-specific rules without exposing more personal information.

Parents and ordinary users should distinguish policy proposals from active law when reading Google News. A governor’s announcement can reveal what regulators want, but only enacted text establishes new statewide duties.

Florida’s experience also offers a broader lesson about AI governance. The most politically compelling protections are often the hardest to implement without new privacy, accuracy, and constitutional risks.

The next successful framework will need more than a list of desirable rights. It must explain who verifies age, who stores the resulting data, who reviews alerts, who can sue, and which government sets the rules.

Will Congress answer those questions before Florida lawmakers try a third time? That is the next action worth following, because the outcome will shape chatbot products far beyond one state.