What you need to know
- Google previewed a new Android feature at Google I/O 2024 that uses AI to detect phrases associated with scams in phone calls.
- Over $12 billion was lost in the U.S. due to digital theft last year, with phishing scams — including fake phone calls — representing the most number of reports.
- Since the AI call detection feature uses on-device AI, there aren't any major privacy implications to using it.
Google I/O 2024 was the most developer-focused developer conference we've seen in quite a while, and it was full of AI talk. The phrase was mentioned in some form or another more than 120 times, in fact. For consumers, most of the AI features mentioned won't have a tangible impact on the way they use technology. That's with the notable exception of one feature — AI scam call detection — that could become the most important Android feature added to the platform in years.
Google is testing scam detection alerts. Although the company shared a preview of the feature in action at Google I/O, we don't have a release date or many details at the moment. Google promised to share more details "later this year," and that's the best timeline we have. Still, the idea behind the feature is a great one, and it looks impressive in demos.
We're testing a new feature that uses Gemini Nano to provide real-time alerts during a call if it detects conversation patterns commonly associated with scams. This protection all happens on-device so your conversation stays private to you. More to come later this year! #GoogleIO pic.twitter.com/l87wGCz62xMay 14, 2024
The proposition is simple: AI will help unknowing users avoid falling victim to a cell phone scam. It uses on-device AI to detect words and phrases commonly associated with phishing scams and alert the user that the call they're receiving is likely a scam. For example, in the demo video, Google showed off a scam detection alert that appeared after a so-called "bank representative" asked a user to transfer funds to a new account.
Essentially, the AI model would be trained to figure out what kinds of requests have a tendency to be fraudulent. The most common type of phishing scams today involve scammers asking people to pay them through gift cards. As such, AI scam call detection could inform a user that a scam is likely once the caller asks them to send gift card redemption codes.
Surprisingly, the feedback coming from users on social media about AI scam detection hasn't been overwhelmingly positive. However, this feature has the potential to do the most good of anything announced at Google I/O, and here's why.
People are losing billions of dollars to scammers
First, let's address the obvious question: do we really need AI to tell us that we shouldn't transfer our life savings based on one unsolicited phone call? It might be hard to believe, but the answer is yes. Chances are, if you're reading Android Central, you probably have a higher understanding of tech literacy and internet safety than the average person. Technology enthusiasts can spot scams instantly, and most of us know better than to even pick up the phone when we don't know the number and aren't expecting a call.
Not everyone is well-versed enough in technology to spot scams quickly and effectively. Older people often fall victim to these types of scams more often than anyone. We can all probably think of someone in our lives who might be fooled by a scam phone call, especially as scammers evolve and find new ways to try to trick people.
If you still aren't convinced spam calls are a massive problem in 2024, here are some numbers to back it up. In the blog post announcing this feature, Google referenced a study from the Global Anti-Scam Alliance. It found that just in 2023, people around the world lost an estimated $1.026 trillion to scams and identity theft. The report included 49,459 people from 43 countries and shows just how prevalent scams are.
Though these are estimated figures, we have more concrete data in the U.S. The FBI releases an Internet Crime Report each year chronicling the reports of online theft and scams and the amount of money that was stolen. In 2023, there was a whopping $12.5 billion lost due to digital scams in the U.S. That's up from $10.3 billion in 2022 and $6.9 billion in 2021. The number of online scams and the amount of money lost are both sharply increasing. For reference, only $3.5 billion was lost in 2019 — so the amount of money lost has tripled in a five-year period.
There are all different kinds of online scams and identity thefts that contribute to the Internet Crime Report. However, the largest number of complaints reported to the FBI came from phishing scams. In 2023, there were 298,878 reports of phishing scams in the U.S., and that's more than all the reports from the other four categories combined.
For those unfamiliar, scams using the phishing method try to trick users into willfully giving up information or money. When someone calls you pretending to be a bank or the IRS, for example, they are attempting to phish your credentials or cash. These are precisely the types of scams that Google's scam detection feature could thwart using AI.
Google won't be listening to your calls — it's all processed on-device
Google's feature works by analyzing your calls in real-time and matching words and phrases with those commonly used in scams. Sounds like a privacy nightmare, right? Actually, it isn't because Google is using on-device processing for the whole thing. For years, companies have used on-device processing to keep certain sensitive types of information from leaving your device. Google has the Titan M2 security chip, and Apple has the Secure Enclave.
In this case, Google is using the Gemini Nano model to detect potential scams in your calls. That means Google won't be able to listen to your calls or even collect any data from them. It'll all happen on your phone, using Google's smallest AI model, without any information being sent down to the company.
This does come with its downsides, like phone compatibility. Since it uses Gemini Nano, the scam call detection feature would likely only be compatible with certain phones like the Pixel 8 Pro and the Galaxy S24 series if it launched today. Gemini Nano is said to be coming to the Pixel 8 later on and could be coming to more phones in the future. But for the sake of privacy, only devices capable of running Gemini Nano will be able to make use of this functionality.
Scam detection alerts will save so many people from theft
I've long said that AI will be far more beneficial to the casual, uninformed tech users than the enthusiasts — especially in the beginning. A perfect example of this concept is Google's plan for AI scam detection alerts. They might not help you or me, but they will help tons of people avoid scam attempts before they are put in an awful situation.
These are the kinds of features that can be the difference between a life-altering mistake being made and a crisis being easily avoided. That's why, throughout all the AI nonsense Google went on about at I/O, these scam detection alerts are what I'm most looking forward to seeing on Android phones.
