Harvard University Warning: 1 in 2 people are deceived… Professional-grade “AI Scam Emails”: Clever Techniques and 5 Defense Techniques

If you think you would never fall for a spam e-mail with suspicious Japanese language, you may have already become a “prime dupe” for the latest scam.

The tremendous evolution of artificial intelligence (AI) is making headlines every day. Its capabilities are so great that there is no end to the concerns about its misuse. In fact, governments and major global companies are strengthening cooperation to prepare for cyber-attacks that exploit the latest AI, both at home and abroad.

While the convenience of AI is of great benefit to society, there is always the risk of abuse.

In this connection, there is an interesting report that sounds a warning bell in our daily lives: “Don’t be fooled by AI scam emails. The report, titled “Don’t Be Deceived by AI Email Scams,” strongly warns that “Harvard experiments show that one in two people will be deceived in the future.

The author of the report, Yu Kashiwamura, chief researcher at Dai-ichi Life Asset Management & Economic Research Institute, said, “We have heard of suspicious e-mails such as, ‘Congratulations on your win! are a thing of the past, and now they have become so sophisticated that it is difficult to tell the difference between legitimate and fraudulent e-mails.

It is astonishing!　Half of all AI scams deceive

First, let’s take a look at the results of an experiment in which a research team at Harvard University in the US used AI to confirm the danger of fraudulent e-mails. In this experiment, 101 people were targeted and randomly divided into four groups and sent emails. They then looked at the percentage of people who clicked on the dangerous links attached to the e-mails.

The four groups were sent (1) conventional junk mail, (2) mail created by experts (professional humans), (3) fully automated mail created by an AI, and (4) mail created by an AI and modified by humans.

As a result, the percentage of those who clicked on the dangerous links was 12% for (1), 54% for (2) and (3), and 56% for (4). In other words, AI scam e-mails are as sophisticated as professional scammers, and we are already entering an era in which one out of two people are fooled.

A closer look at the results of the survey reveals a frightening fact. The subjects of the experiment were university faculty across a variety of disciplines, including the natural sciences. These people are considered to have a higher level of information literacy (reading and writing ability) than the general public. Moreover, the survey was conducted two years ago. There have been remarkable technological advances in AI over the past few years, including the appearance of “Claude Mutos,” mentioned at the beginning of this article.

Sophisticated methods to exploit weaknesses from social networking sites

Mr. Kashiwamura, aforementioned, describes the transition of fraudulent e-mails as follows: “Before, many people would send the same e-mail to many people at once.

In the past, the same e-mail was sent to a large number of people in large quantities, with the possibility that some of the many people who received it might be fooled.

What does this mean?

AI collects information, organizes it, and makes sophisticated inferences.

As Mr. Kashiwamura says, the current AI automatically searches the Internet for the target person’s name. It then thoroughly collects publicly available information, such as posts on Facebook and other social networking services, self-introductions on company websites, and past news articles.

Next, the information is analyzed to determine what kind of work the person does, what his/her interests are, and what kind of activities he/she is involved in. Finally, they compile a detailed profile of the person’s gullibility, weaknesses, and believability points.

The Harvard team also interviewed the experiment participants about why they believed the AI scam emails were genuine. As a result, they found that the content had become so sophisticated that it was easy for anyone to believe, with comments such as, “Because it was relevant to me,” and “It was a special e-mail addressed to me.

For example, after mentioning the name of a project in which the target had recently been involved and his area of expertise, the following e-mail would arrive

We would be very grateful if you could help us with this matter.

With this kind of text, it is extremely difficult to be skeptical.

Furthermore, sophisticated AI scam e-mails are now sent in fluent Japanese, even from overseas. In the past, it was easy to spot a scam by the presence of awkward Japanese or simplified Chinese characters, but these are no longer the case.

Must Read!　Five Defenses Against AI Scams

Most of these fraudulent e-mails are related to money, such as solicitation for investment management.

They exploit human greed and psychology.

Mr. Kashiwamura points out this and lists “5 measures” to protect oneself from fraudulent e-mails.

Five measures to protect yourself

• 1. Be suspicious of e-mails from people you know: Be wary of e-mails from people claiming to be your parents, siblings, or spouse, even if they say they need a large sum of money right away.
• 2_Do not click on links or buttons: There is a high possibility that a trap is lurking that will lead you to a fake site that looks just like the real thing.
• 3_Use official channels for confirmation: Do not use the contact information in the e-mail, but always confirm through official sites or other legitimate channels.
• 4_Review public settings on social networking sites: Protect yourself by limiting the scope of publicly available personally identifiable information, such as your place of employment, job title, recent places visited, hobbies, etc.
• 5_Set up multi-factor authentication: The level of authentication should be increased by allowing multiple methods of identification, such as authentication codes and biometrics, in addition to passwords.

Voice forgery?　The Terrifying Evolution of AI

Recently, AI has come to accurately imitate even the human voice, and even fluent speech.

I am afraid of voice now.

As Mr. Kashiwamura says, the risk of receiving a fake phone call from someone pretending to be a relative or other acquaintance and believing the caller’s voice is real is now a reality. In closing, Mr. Kashiwamura sounded the following warning about AI and other advanced technologies and the current state of Japanese people.

Many Japanese people have low IT literacy. It would be better to have them study AI properly from the elementary and junior high school stages.