How the use of AI and ‘deepfakes’ will play a role in the search for Nancy Guthrie: NPR

The search for Nancy Guthrie, who disappeared from her Tucson, Arizona, home two weeks ago, has continued and fake kidnappers have swarmed.

Law enforcement said they received several ransom notes from people claiming to be the child’s mother. today show co-host Savannah Guthrie; Savannah and her siblings, Annie and Camron, posted a video on social media pleading for 84-year-old Nancy Guthrie to return home, demanding proof of her survival before a ransom is paid.

“We are ready to speak, but we live in a world where voices and images are easily manipulated,” Savannah Guthrie said in a video, asking for proof that her mother was still alive.

Law enforcement says as artificial intelligence becomes more sophisticated and commonplace, it can become difficult to determine what is real and what is fake, complicating the search for Nancy Guthrie. But how difficult is it?

AI can “create almost anything”

Before the advent of artificial intelligence, proof of survival could easily be established by having a hostage take a photo while holding the day’s newspaper or talking on the phone, said Joseph Lestrange, who worked in law enforcement for 32 years and now trains law enforcement agencies to identify artificially generated content.

Now, someone can ask a language learning model to imitate someone’s voice or likeness in a photo, video, or audio, known as a “deepfake.” Lestrange said models can also create fake documents, such as passports.

“If you give them the right instructions, they can do almost anything,” Lestrange said.

In federal agencies, digital evidence is typically sent to a digital forensics lab. There, examiners can use clues such as location data and pixels to determine the authenticity of evidence. Lestrange said the tools they are using are “very effective,” but it will take time to draw conclusions.

“Typically time is of the essence in these kidnapping cases, especially in this case we’re talking about, where the poor woman has health issues,” he said. “So these investigators are in a really difficult situation right now.”

Lestrange said local and state agencies may not have access to similar tools, while fraud is becoming more complex and swift.

Lestrange said some government agencies are more open to using artificial intelligence than others. This problem could begin to be fixed if emerging AI companies work with law enforcement agencies to “develop meaningful products,” he said, so law enforcement agencies aren’t “just relying on vendors to tell them what they need.”

How to protect yourself from AI fraud and deepfakes

While artificially generated content and deepfakes rely on digital tools, human intervention and judgment is still a good way to know if something is wrong, said Eman El-Sheikh, vice president of the Cyber ​​Security Center at the University of West Florida.

“First, stay calm and slow down, because scammers often try to create a false sense of urgency to get their way before others realize it’s a fake,” she says.

During the call, you can say things that you know will likely cause your loved one to respond in a certain way. Alternatively, you can hang up and call your loved one directly to check on the problem, El-Sheikh said.

People who use social media should avoid disclosing sensitive information such as passwords, addresses and phone numbers, he said. It’s also important to keep details about your home private, such as when you leave the house or whether you live alone.

Also, be sure to check your app’s privacy settings and toggle permissions based on your comfort level, she said.

“It is extremely important for everyone living in a digital world to be intentional about protecting the information and privacy they say online.”

But Lestrange points out that anything you post or share about yourself online can be used against you, even if you’re careful.

“The world today is really quite different,” he said.