Florida Attorney General Proposes New Powers to Shut Down Scam Calls and AI Voice Fraud

Florida Attorney General James Uthmeier is proposing legislation that would give his office independent authority to investigate and shut down spam call and text operations, and would create tougher penalties for a set of specific deceptions including impersonation scams, fake warrants, and undisclosed use of artificial intelligence voices. The proposal, outlined in mid-September, is aimed at the next legislative session.
The AI voice provision is the newest element and the one that distinguishes this proposal from previous rounds of telemarketing legislation. Voice cloning technology that once required substantial technical resources is now available cheaply and requires only a short sample of someone speaking. That capability has moved a category of fraud from theoretical to routine.
Florida has particular exposure. The state has one of the largest populations of older adults in the country, and older adults are disproportionately targeted by phone-based fraud. Florida is also a retirement destination, which means a large number of residents have adult children and grandchildren living in other states, the exact family structure that grandparent scams exploit.
What the proposal would do
The first component is enforcement authority. Under the proposal, the Attorney General's office would have independent authority to investigate and shut down spam call and text operations. The significance of the word independent is that it would reduce the office's dependence on referrals or on coordination with other agencies before acting.
The second component is penalties. The proposal would create tougher penalties for impersonation scams, a category covering callers who claim to represent a government agency, a utility, a bank, or a law enforcement office. Fake warrant scams are called out specifically. Those operations tell a target that a warrant has been issued and that payment will resolve it, and they are effective because most people do not know how warrants actually work.
The third component addresses undisclosed artificial intelligence voices. The framing centers on disclosure, which is a narrower and more legally durable approach than attempting to ban synthetic voices outright. A rule requiring that a caller disclose when a voice is machine-generated targets the deception rather than the technology.
Together the three pieces describe a strategy of giving the office faster tools against high-volume operations while raising the cost of the specific tactics that cause the most financial harm.
Why existing law has not solved this
Telemarketing fraud has been illegal for decades, at both the federal and state level. Federal law restricts automated calls and prerecorded messages, the Federal Communications Commission and Federal Trade Commission enforce against violators, and Florida has its own telemarketing statutes and a state do-not-call list.
The enforcement problem is structural rather than legal. High-volume scam operations are frequently based overseas, beyond the practical reach of state subpoenas. They use voice over internet protocol services that allow caller identification spoofing, so the number a recipient sees bears no relationship to the actual origin. They operate through layers of intermediary carriers, which makes tracing a call back to its source a technical investigation rather than a records request.
Scale compounds it. A single operation can place millions of calls, and the economics work even at very low response rates. Enforcement actions that take months to build are outpaced by operations that can relocate their infrastructure in days.
Industry and regulatory efforts have made progress on authentication. Caller identification authentication frameworks adopted by carriers have reduced some spoofing, and carriers now block substantial volumes of suspected fraud calls. None of that has eliminated the problem, in part because scammers have shifted toward text messages, which have different technical characteristics and arrive with fewer filters.
How AI voice fraud works
The mechanics are simpler than most people expect. A voice cloning tool needs a sample of the target voice, and social media video, voicemail greetings, and podcast or interview audio all supply usable samples. From that, the tool can generate speech saying anything.
The classic application is the family emergency scam. A parent or grandparent receives a call in what sounds like their child's or grandchild's voice, in distress, describing an accident, an arrest, or a hospital, and asking for money urgently and in secret. The emotional pressure is the mechanism. Voice cloning makes the pretext far more convincing than a stranger claiming to be a relative.
The same technology serves business fraud. A cloned executive voice authorizing a wire transfer defeats the informal verification that many organizations rely on, which is that someone recognized the voice on the phone. Finance staff trained to confirm unusual transfers by calling the requester can be defeated when the call itself is synthetic.
Verification is the practical defense, and it does not require technical sophistication. Hanging up and calling the person back on a known number defeats nearly every version of this scam, because the attacker does not control the number being called. Families who agree on a code word in advance have a second layer that works even under emotional pressure.
Why Florida is a target
Florida's demographic profile makes it attractive to phone fraud operations for reasons that are entirely mechanical. A large population of retirees means a large population of people who answer landlines, are home during business hours, and hold accumulated savings.
The geographic separation of Florida retirees from their families amplifies the family emergency scam specifically. A grandparent in Naples whose grandchildren live in Ohio has no immediate way to check on a claimed emergency, and the distance itself makes the story plausible.
Florida's real estate market contributes a separate category. Wire fraud targeting real estate closings has been a persistent problem, with attackers inserting themselves into transactions and redirecting closing funds. Florida's transaction volume and its high share of cash buyers make it a productive environment for that fraud, and a cloned voice authorizing a change in wire instructions is a natural evolution of it.
The state's disaster exposure produces yet another. After hurricanes, fraudulent contractor and insurance claim operations proliferate, and phone-based solicitation is one of their channels. Any expansion of enforcement authority against high-volume calling operations has application in post-storm periods.
The AI policy context in Tallahassee
This proposal arrives against a backdrop of Florida activity on artificial intelligence. Governor Ron DeSantis previously amended a special session call to include an artificial intelligence bill of rights, framed around protecting Floridians, and particularly minors, from deceptive practices by large technology companies. Bills were filed in the Senate, but the House did not file companions, and the issue was not taken up once the Legislature convened.
That history suggests the political appetite exists but that the legislative vehicle has not yet come together. An Attorney General proposal focused narrowly on fraud may have an easier path than a broad AI rights framework, because consumer fraud enforcement attracts less industry opposition than general regulation of AI systems.
Uthmeier has also been building a record on the enforcement side of AI. He announced charges in September against a man accused of sexually abusing and recording a child, and used the announcement to describe the state's expanding use of artificial intelligence and undercover operations to identify child predators. That is AI as an investigative tool rather than a subject of regulation, and it shows the office engaging the technology from both directions.
Uthmeier has separately announced an audit of public spending, indicating an office pursuing an active agenda across several areas ahead of the session.
What it would mean for Floridians
For a resident, the most tangible potential effect is fewer calls, though experience counsels modest expectations. Every previous round of telemarketing legislation has been followed by adaptation rather than elimination, and operations based outside the United States remain difficult to reach regardless of what authority a state grants its attorney general.
The penalty provisions matter most for operations that touch Florida directly, including domestic call centers and the payment processors and money transmitters that convert a successful scam into cash. Enforcement against that infrastructure is often more productive than pursuing the callers themselves.
The disclosure requirement for AI voices would give the state a specific, provable violation. Proving intent to defraud is difficult. Proving that a call used a synthetic voice without disclosure is comparatively straightforward, which is exactly why disclosure rules are attractive to enforcers.
What Floridians can do now
None of this legislation exists yet, and the defenses available today do not depend on it.
- Hang up and call back on a number you already have, for any call claiming an emergency or demanding payment
- Treat any demand for payment by gift card, cryptocurrency, or wire transfer as fraud, because legitimate agencies do not request those
- Agree on a family code word for emergencies, which defeats voice cloning
- Verify wire instructions for any real estate or business transaction by phone using a number from your own records
- Register on the state and federal do-not-call lists, and use carrier-provided call blocking
- Report fraud attempts to the Florida Attorney General's office, since enforcement priorities follow complaint volume
Reporting is more consequential than it appears. Individual complaints are what allow investigators to establish patterns, identify infrastructure, and build cases against operations that would otherwise look like isolated incidents.
What's next
The proposal has to become a bill, find sponsors in both chambers, clear committees, and pass before it means anything. The history of the artificial intelligence bill of rights, which stalled when one chamber did not file a companion, is a reminder that a governor's or attorney general's endorsement does not guarantee passage.
The details that will matter in the bill text are the scope of the independent investigative authority, whether the AI disclosure requirement covers texts and messaging as well as voice calls, and what penalties actually attach. Those specifics will determine whether the legislation is a meaningful enforcement expansion or a restatement of existing prohibitions.
Floridians who want to weigh in have the ordinary avenues: contacting legislators, following the bill once filed, and participating in committee comment. Consumer protection bills generally attract less public attention than their practical effect on households would justify.
Spotted an issue with this article?
Have something to say about this story?
Write a letter to the editor
