FCC Drops AI Voice Ban On Robocalls

The Federal Communication Commission (FCC) announced on February 8 a ban on robocalls containing voices generated through artificial intelligence, the Associated Press reported.

The FCC’s unanimous ruling bars the use of AI voice-cloning tools in robocalls under the 1991 Telephone Consumer Protection Act, the law that restricts spam calls using artificial or prerecorded voice messages.

The decision comes as authorities in New Hampshire investigate AI-generated robocalls featuring the purported voice of Joe Biden urging state residents not to vote in the Democrat primary in January.

The new rule will allow the FCC to fine any company that employs AI-generated voices in their calls and block any service provider that would carry them. It would also pave the way for recipients of the AI-generated robocalls to file lawsuits against the companies and empower state attorneys general to crack down on those who violate the rule.

FCC Chair Jessica Rosenworcel told the Associated Press that bad actors use AI-generated voices to misinform voters, impersonate people, and even extort family members.

Under existing laws, telemarketers are generally not permitted to use auto-dialers or prerecorded or artificial voice messages when calling cell phones and can only make calls to landlines with prior written consent from the recipient.

With the new FCC rule, AI-generated robocalls are classified as “artificial” and therefore the same standards would apply. Violators would face fines of as much as over $23,000 per call.

Aspen Institute’s Josh Lawson, the director of AI and Democracy, told the Associated Press that even with the ruling, voters should be prepared to receive personalized, targeted messages via phone calls, texts, and social media. He said bad actors “know what they’re doing is unlawful” but they will “continue to rattle the cages and push the limits.”

Carnegie Mellon University professor Kathleen Carley, a specialist in computational disinformation, told the Associated Press that currently, it is possible to identify audio generated by artificial intelligence. But as the technology improves, that may become increasingly difficult.