How tech companies in Mass. are trying to guard against deepfakes

Artificial intelligence can generate very convincing copies of peoples' voices.

For example, last month, Republican gubernatorial candidate Brian Shortsleeve used an AI generated copy of Gov. Maura Healey's voice in an ad attacking her and lawmakers on Beacon Hill are scrambling to create guardrails around AI in political communications.

In the tech world, companies are also emerging to help detect AI-created media and fraud.

Scott Kirsner, a columnist with our editorial partner MassLive, joined WBUR's Morning Edition to discuss.

Highlights from this interview have been lightly edited for clarity.

On what local companies, like Modulate AI, are doing to combat AI fraud:

"Initially they were in the world of video gaming. If you play some of these video games, like Call of Duty or Grand Theft Auto, there is live chat where you can talk to other players. Abuse and hate speech and, to some extent, child grooming can be a problem in these live chat spaces. And so Modulate basically said, 'Hey, we're going to suss out when we think those things are happening by listening to the voices and analyzing the intonation, the intent. And we're going to be able to tell the difference between something that's really threatening and something that is maybe benign trash talk.' "

On other players in this space in Massachusetts: 

" Modulate is the main one that I've seen in this voice AI detection space. Here in Massachusetts, there are two companies: One's called Reality Defender, looking at fake video and fake audio. I love the name of that company, and they have an early investor that's here in Massachusetts. There's another one called Resemble AI, which is kind of like trying to sell both the guns and the bulletproof vests. They let you make voice clones with their software, and they also will then sell detection capability."

On how private companies are helping cut down on AI-related fraud:

"If you think about some of the places that fraud happens right now, it's in call centers, where someone might call your bank with a synthetic AI voice and pretend to be you and do something like change your password, which would be really bad. And so companies like Modulate are trying to develop software, they're trying to adapt their software to say, if you're in a call center, we're going to do two things.

"We're going to first tell you if we think this is a synthetic voice rather than a real human being calling you. ... The second thing that they want to be able to do is understand what kind of fraud you're trying to perpetrate — like an account takeover where you get a password reset. And they're going to show that to the call-center employee, giving you a yellow light and then a red light that says, 'Hey, this person or this synthetic voice is trying to do something that is against our policies, so don't let them.' "

On what businesses want to see from Beacon Hill legislators:

"I don't know that they necessarily want to see regulation, but they don't want to lose money to fraudsters. I think we're going to really see three things happen.

"One is regulation probably in more states, hopefully Massachusetts too, for having to disclose that you used AI in creating a campaign ad.

"Companies are going to be motivated by losing money to detect fraud — they may not need to be legislated.

"But the third thing, which we haven't talked much about societally yet, is us as individuals. There's not as much money in protecting you and me from a scammer as individuals, and so are we going to see technology — whether it's built into Instagram or it's built into your phone when you pick up a phone call, you probably also want to detect AI voices — people pretending to be who they're not and do nefarious things — and so I hope that becomes a topic of discussion."