Experts Warn Against Government Moving Too Quickly to Regulate AI in Telecom
With the FCC considering an NPRM and notice of inquiry as the agency's next steps on AI, the issue of AI and robocalls will only grow in importance, experts said Thursday during an FCBA webinar. They also agreed that generative AI could yield new tools that can help curb unwanted and illegal texts and calls.
Sign up for a free preview to unlock the rest of this article
Communications Daily is required reading for senior executives at top telecom corporations, law firms, lobbying organizations, associations and government agencies (including the FCC). Join them today!
“There’s a lot going on with AI,” said Wiley’s Kevin Rupy, who represents clients on robocall issues. AI is “currently the shiny object ... drawing tremendous interest from both industry and government stakeholders,” he said. For example, there are rulemaking proceedings at the FCC and FTC as well as enforcement actions from both agencies. Other agencies, including DOJ, DOD and the Department of Homeland Security, are also focused on AI, as are some states, he added.
The biggest problem with AI calling is people often don’t know whether they’re talking to a human or a machine, said Vijay Balasubramaniyan, CEO of Pindrop, which focuses on deep-fake detection and voice security. “We’re seeing some crazy tactics by fraudsters who are using AI … to do all kinds of nefarious things,” he said, adding, a “tsunami is coming.”
For bad actors, “regulation doesn’t matter and that’s where you need great detection tools,” Balasubramaniyan added. “We’re in the early innings of AI, and if you try to regulate something that’s coming so quickly, you’re going to lose what makes it great.” Regulators need to give AI time to develop to determine what regulation can do to help consumers, he said.
Balasubramaniyan said that ultimately a robocall is a “piece of audio.” When companies use voice-generative technology “they leave behind telltale artifacts of the fact that it was machine-generated." When you say a simple phrase, “your entire vocal cavity, nasal cavity, your diaphragm, your vocal cords, go through a certain set of configurations ... at a certain speed.” He continued, “Machines don’t care about any of that.” Every second, a machine has 8,000 opportunities to make a mistake, he said.
Consumers love their wireless phones “and we use them all day, every day,” said Sarah Leggin, CTIA assistant vice president-regulatory affairs. The wireless industry is focused on protecting consumers from bad actors “but we know that there’s always more work to do,” she said. “AI can be a way that existing robocallers and existing robotexters scale and enhance their attacks.” Carriers are already using different tools to protect customers from “tons” of unwanted robocalls and robotexts.
AI has been around since the 1950s, while generative AI really got started around 2017, said Mike Rudolph, CTO at YouMail, a visual voicemail and robocall-blocking company. Millions of companies have been using non-generative AI in some way, he said. Typical uses are analyzing data, “getting insights ... finding a pattern in the data, classifying the data.” One example is spam tools that have been available for use with email for more than a decade, he said. Generative AI is “a scary boogeyman” but also “a fantastic super tool” to analyze data and communications in a more sophisticated way.
The FCC had to clamp down on voice-cloning technology in robocall scams last year (see 2402080052) because AI-based technologies have made it harder to distinguish an artificial voice from a real one, said Edyael Casaperalta, aide to FCC Commissioner Anna Gomez. “What we’re trying to do is protect consumers from this emerging new technology,” but if a caller really wants to use AI, the FCC says that’s also OK, “just make sure that you get prior and expressed consent from the individual you’re trying to call.”
The NPRM/NOI, approved 5-0 in August (see 2408070037), asks “at a high level … how we need to account for [AI] in the robocall and robotexting space,” Casaperalta said. “We are concerned with the harms that could happen, but also looking at ... the possibilities.” The notice explores whether the FCC should have a definition for an AI-generated call. There is already much discussion in the record and the FCC wants to hear from affected parties, she added.
In addition, the FCC is looking at the need for an exemption for people with disabilities using AI, Casaperalta said. “We’re very interested in that and so, if there’s a use for AI that can help this community, we don’t want to foreclose it.”
The FTC ran a contest to award innovation, particularly among companies that block unwanted robocalls, said Benjamin Davidson, attorney in the Division of Marketing Practices in the Bureau of Consumer Protection at the agency. “We believed there needed to be technological developments that we didn’t see happening,” he said. One of the winners, Nomorobo, is now used by many people to block calls, he said.