Chapter 5. Recognizing AI Scams and Fakes, and Everyday Uses

Learning objectives

• Explain what a deepfake (a fake that imitates a voice or face) is, and why it is difficult to distinguish by listening and seeing alone.
• Identify common signs of scam communications (urgency, secrecy, demands for money and financial information, link/app installations, unfamiliar numbers and messengers).
• Follow the step-by-step procedure of hanging up, verifying with a saved number, and reporting when receiving suspicious communications.
• Explain how to establish and use verification questions and emergency contacts among family members in advance.
• Explain tips for being suspicious of fake photos, news, and reviews, as well as the meaning and limitations of labels on AI-generated content.
• Use AI safely for small everyday tasks such as summarization, translation, and writing.

Introduction

At 11:00 p.m., Ok-sun Yoon's mobile phone rang. It was an unknown number. As soon as she answered, she heard a tearful voice. "Mom, it's me... I got into an accident. Mom, what do I do?" It was unmistakably her daughter's voice. Even the sound of her breathing was identical. Immediately, an unfamiliar man took over the call. "Your daughter damaged our car. If you send a settlement of KRW 5,000,000 right now, we won't report it to the police. If you hang up, your daughter will get hurt. You must not tell anyone else."

Ok-sun Yoon's hands trembled. Just as she was about to open her banking app, she remembered a promise made with her family last Lunar New Year. "If money is mentioned, hang up unconditionally, call back using the saved number, and ask one question we decided among ourselves." Ok-sun Yoon said, "Hold on, I'll go to the bank," and hung up. Then she dialed her daughter's number saved in her phone. Her daughter answered in a half-asleep voice. "Mom, what's going on? I was sleeping."

That voice was very likely a fake created with AI. These days, videos or audio posted on social media can be used as raw material to imitate someone's voice. Distinguishing real from fake just by listening is becoming increasingly difficult. Therefore, this chapter teaches "the procedure to prevent damage even when it is fake" before "the eye to spot fakes." First, we look at what deepfakes are, and learn the common signs of scam communications, verification procedures, family verification questions, and places to report. Next, we cover fake photos, news, reviews, and labels on AI-generated content, and finally conclude by bringing together what we learned in this book to use AI safely for small tasks in everyday life.

Key concepts

What is a deepfake?

A deepfake refers to fake photos, videos, or audio created by using AI to imitate a person's face or voice. It is a compound of "deep learning," a branch of AI technology, and "fake." Imitations that replicate only voices are sometimes separately called deepvoice.

Deepfakes appear in forms like these:

To create a deepfake, the face or voice of the person to be imitated is required. That material usually comes from public places. When providing guidance on phishing that imitates family members' voices using AI, the National Police Agency urged people to be careful when posting content that includes audio on social media, and regarding fake Child abduction videos made with deepfakes, also advised people to avoid posting with public visibility settings that anyone can view.

Why it is difficult to distinguish by listening and seeing alone

In the past, one could recognize scam calls because the voice was awkward or the dialect was strange. Today, even a person's speaking style and breathing sounds can be imitated. The same applies to fake videos. While some advise looking for flaws such as blurred face edges or misaligned mouth movements, these flaws decrease as technology improves. Furthermore, hearing a family member crying in the middle of the night prevents anyone from calmly assessing the situation.

Therefore, this book recommends a different direction:

• Do not judge with your ears and eyes. "The voice sounded real" is not evidence.
• Verify through procedure. Hang up, call back using a saved number, and ask predetermined questions.
• The procedure is useful even in real emergencies. If an accident might really have occurred, contact the person involved or other family members using a saved number to confirm their safety, and report it to 112 at the same time. If it is a real accident, help from the police and paramedics is needed; if it is fake, this procedure protects your money. Either way, sending money first to a stranger's account is not the answer.

Five signs of scam communications

Whether AI is used or not, scam communications have recurring signs. If even one of the following five signs appears, stop. Even if the signs are not clear, if something feels strange, stop and verify separately.

Among the signs, method 4—sending a link via text message to direct victims to a fake website or install a malicious app—is separately called smishing. It is a compound of SMS (short message service) and phishing. Never click links inside text messages disguised as package tracking, notifications of family events (weddings or funerals), or fine notices.

Guidance from official institutions highlights the same signs:

• The Financial Supervisory Service Voice Phishing Keeper advises to never respond when account numbers, card numbers, or internet banking information are requested over the phone or internet, and that luring you to an automated teller machine (ATM) under the pretext of tax refunds or account protection is 100% voice phishing.
• The National Police Agency explains that claiming to be a prosecutor, telling you not to tell anyone, and intimidating you with an investigation is a phishing technique intended to create fear and prevent reporting.
• The Financial Supervisory Service advises that even if the other party approaches you already knowing your personal information (Privacy) or financial transaction information, you must verify the authenticity of the phone call, text message, or messenger content. Knowing your name and address does not mean it is real.

Stop, hang up, and verify

This is the procedure to follow when you receive suspicious contact. Memorize the order:

1. Stop. Do not send money, read out numbers, click links, or install apps.
2. Hang up. Saying a single sentence is enough: "I will check and call you back." Hang up even if the other person tells you not to hang up.
3. Call directly using a saved number. If it is family, call the number originally saved in your mobile phone; if it is an institution, call the main telephone number listed on its official website.
4. Ask the predetermined verification question. If there is a question agreed upon in advance among family members, verify with it (next section).
5. Notify other family members as well. Do not judge alone. Other family members can verify the person's safety together.
6. Report. If it appears to be a scam, report it; if you have already sent money, contact the financial company and the police immediately (see "Where to report and get help" in this chapter).

The National Police Agency also advises that when suspicious, hanging up the phone first and checking with the person directly or with acquaintances around them is the fastest and safest prevention method.

Verification calls to a saved number

What matters in Step 3 is which number you dial. If you make a "verification call" to a number given by the other party, that number may belong to the scammer. The same applies to numbers received via text message.

The Financial Supervisory Service advises to note that caller ID can be manipulated. This means that even if a bank's or institution's number appears on the screen, or even if it looks like a family member's number, you cannot trust it as real based solely on that. Therefore, the rule is not to verify over the incoming call, but to hang up and call directly yourself.

When it is difficult to hang up

When threatened with words like "If you hang up, your daughter gets hurt," it is frightening to hang up. The National Police Agency provides this guidance for such cases: If it is difficult to report because the criminal threatens you not to hang up, ask people nearby for help to report to the police, or report to 112 via text message while on the call to confirm your child's safety. The National Police Agency emphasizes that whether it is an abduction or a scam disguised as an abduction, you must report to the police unconditionally.

Preparing in advance in everyday life helps in these moments:

• Save family members' mobile phone numbers as speed dials or favorites.
• The Financial Supervisory Service advises to secure contact information for your child's friends, teachers, relatives, and others in advance to prepare for Child abduction voice phishing. This is because there must be an alternative way to verify safety when the person involved does not answer the phone.

Setting family verification questions

A predetermined verification question among family members is a simple and powerful way to counter scams that imitate voices and faces. While scammers can imitate voices, it is difficult for them to know secrets known only among family members.

The conditions for a good verification question are as follows:

How to establish them:

1. Decide on them in an in-person meeting, such as during holidays or family gatherings. If decided in group chats or text messages, those records could leak.
2. Do not write the questions and answers on mobile phones or social media. If you must write them down, write them on paper and keep it in a safe place.
3. Establish the rule that "if money is mentioned, unconditionally hang up and call back using the saved number" as well. This rule comes before verification questions.
4. Talk about it again about once a year so it is not forgotten.

Family impersonation via messenger

There are also many impersonations that arrive via messengers rather than phone calls. Examples include messages such as: "Mom, my phone screen broke so I'm contacting you from a friend's phone. I have an urgent payment to make—could you send a photo of your card and your password?" They may even customize their profile picture and name to look like family.

The five signs are present in these messages as well: it is urgent (1), from an unfamiliar number (5), and demands financial information (3). As guided by the Financial Supervisory Service, always verify the authenticity of messenger contents. Asking "Are you really [Name]?" via messenger is not verification. The scammer will merely reply, "Yes, it's me." Call the originally saved phone number to verify by voice, and with the verification question. If you cannot reach the original number, verify through other family members.

Fake photos and videos

Photos and videos created with AI are very plausible. They can create accident scenes that never happened, statements celebrities never made, and pictures of items that differ from reality. The thought that "it has a photo, so it must be real" is no longer valid.

When encountering fake photos and videos, look at the source rather than the image itself:

1. Who posted it first? Is it an official institution, a news organization, or the individual's official account, or is it a forwarded post of unknown origin?
2. Has it been reported by other trustworthy sources? If it is a major event, multiple news outlets and official institutions cover it.
3. What is it telling me to do? If it tells you to send money, invest, click a link, or spread it quickly, be suspicious.
4. Does it strongly stir emotions? The more excessively surprising, scary, or enraging a video is, the more you should pause once again.

In particular, videos of celebrities recommending investments are frequently used in deepfake scams. As seen in Chapter 3, if phrases such as "guaranteed high returns," "principal guaranteed," or "right now" appear, be suspicious regardless of whose face is in the video. Verify whether the celebrity actually said those words through their official channels or press reports.

Fake news and fake reviews

AI can generate plausible text very quickly and in large quantities. Therefore, it is also used to create fake news and fake reviews.

Signs to suspect fake news

• It looks like an article, but has no news organization name, or has a name heard for the first time.
• It lacks a date, or describes an old event as if it happened recently.
• It is framed secretively with phrases like "what the government is hiding" or "what no one is telling you."
• It claims to be an announcement from an official institution, but no such announcement exists on that institution's website.
• It asks you to "please spread this quickly."

Signs to suspect fake reviews

• Sentences across multiple reviews are strangely similar.
• Complimentary reviews are clustered together within a short period.
• It only contains expressions like "the best" or "must-have item" without specific usage experiences.
• It only has excessively good words, with not a single downside mentioned.

The verification habits from Chapter 3 apply directly here as well: check whether it is an announcement from an official institution, when the information was published, and whether it can be confirmed through other reliable sources. And before verifying, do not forward it to group chats.

Labeling AI-generated content

There is a law requiring labeling so that people can know whether text, images, or videos were created by AI. Article 31 of the AI Basic Act stipulates as follows:

• Paragraph 1: When an AI business operator intends to provide products or services using high-impact artificial intelligence or generative artificial intelligence, it must notify users in advance of the fact that they operate based on the artificial intelligence in question.
• Paragraph 2: When an AI business operator provides generative artificial intelligence or products/services using it, it must label the fact that the output was generated by generative artificial intelligence.
• Paragraph 3: When an AI business operator provides outputs such as virtual sounds, images, or videos that are difficult to distinguish from reality using an artificial intelligence system, it must notify or label the fact that they were generated by an artificial intelligence system in a manner that users can clearly perceive. In the case of artistic or creative expressions, this may be done in a manner that does not impair display or enjoyment.
• Paragraph 4: Methods of notification and labeling, exceptions, and other details shall be prescribed by Presidential Decree.

Here, an AI business operator refers to a corporation, organization, individual, state agency, or other entity engaged in business related to the artificial intelligence industry that falls under either an AI development business operator that develops and provides artificial intelligence (Subparagraph a) or an AI utilization business operator that provides products or services using artificial intelligence provided in such a manner (Subparagraph b) (Article 2, Subparagraph 7); a user is a person who is provided with artificial intelligence products or services (Article 2, Subparagraph 8). In short, this obligation lies with the party providing the AI service.

There are two ways to interpret this law in daily life:

Where to report and get help

These are the places to contact when a scam is suspected or damage has already occurred. This summarizes official guidance as of the verification date (October 28, 2026).

If you have already sent money, follow the "How to Respond When Victimized by Voice Phishing" from the Financial Supervisory Service Voice Phishing Keeper. The core points are as follows:

1. Immediately call the customer center of the bank that sent the money or the bank holding the receiving account to report the damage and apply for suspension of payments (account freeze) on the account. A payment suspension is a measure that locks the scammer's account so that money cannot be withdrawn.
2. If your personal information (Privacy) has leaked or you installed a malicious app, take measures such as resetting your mobile phone, deleting the malicious app, registering with the Financial Supervisory Service's Personal Information (Privacy) Exposure Accident Prevention System, suspending payments on accounts via the Korea Financial Telecommunications & Clearings Institute's Account Information Integration Management Service, and checking whether mobile phones were opened in your name through the Korea Association for ICT Promotion's Identity Theft Prevention Service.
3. Obtain a Confirmation of Incident/Accident Facts from the police, and submit a written application for relief of damage to the bank branch where you applied for payment suspension. If you applied for payment suspension by telephone, you must submit application documents within 3 business days from the date of application (Article 3, Paragraph 2 of the Enforcement Decree of the Special Act on the Prevention of Loss Caused by Telecommunications-based Financial Fraud and Refund for Loss; FSS guidance states "within 3 business days of the application date"). If this period expires, the bank specifies a period of 14 days and notifies you to submit the documents; if you still do not submit them within that period, the application made by telephone is deemed not to have been made (Paragraph 3 of the same Article). Inquire with the bank or police station regarding required documents before visiting.

Everyday use 1: Summarization

Now, we bring together what we learned in this book to see how to use AI in daily life. First is summarization. AI is very good at shortening difficult-to-read text, such as apartment management office announcements, school newsletters, and long notices.

Try this: Extracting only action items from a long notice

1. Erase information such as names, addresses, contact details, and customer numbers from the notice (Chapter 4).
2. Paste the notice and ask like this:

[Prompt to send]
Below is an announcement from our apartment management office.
Please extract only what I need to do and list them as numbered items, up to three.
Transcribe the date, time, and amount exactly as written in the announcement, and do not add any content that is not in the announcement.

(Paste announcement content here)

3. After receiving the AI's summary, cross-check the date, time, and amount against the original text once more. It may alter or omit numbers while summarizing.

Everyday use 2: Translation

AI is also of great help when translation is needed, such as with menus at travel destinations, manuals for foreign products, or greetings to send to foreign neighbors.

There are three tips when using translation:

1. Provide the context. If you state the intended use, such as "a sentence to show to restaurant staff" or "a polite greeting to a neighbor," it translates in an appropriate tone.
2. Back-translate to check. Put the translated sentence into a new conversation and try saying, "Translate this sentence into Korean." If it differs significantly from the original meaning, translate it again.
3. Leave critical translations to humans. For translations where mistakes could lead to major consequences, such as medical treatment at hospitals, contracts, and official documents, use professional interpretation/translation or the official guidance of the institution concerned.

[Prompt to send]
Please write a sentence in Japanese to show to the staff at a restaurant while traveling in Japan.
"I have a peanut allergy. Please let me know if there are any dishes that contain peanuts."
Please write it in a polite tone, and write its meaning below the Japanese sentence again in Korean.
I want to check if the meaning is correct.

For sentences concerning the body, such as allergies, in addition to back-checking the translation result, it is safer to show it once to someone who knows the language before traveling.

Everyday use 3: Writing

Writing messages of congratulations, gratitude, apologies, and requests is one of the things AI does best. The five-cell prompt table from Chapter 2 applies directly here. When using AI for writing, remember the following:

• Input facts yourself. Provide dates, locations, names, and amounts yourself, or ask the AI to leave them as blanks so that it does not fabricate them.
• Revise sentiments to be your own. Sentences from AI are unobjectionable, but they tend to be generic statements that could apply to anyone. Adding a single line of memory known only to you makes it your own writing.
• Write honestly. Do not include untruthful content in second-hand trade posts, reviews, or civil complaints (Chapter 2). Do not embellish items with AI-generated photos to make them look better than they actually are.
• Read before sending. Read aloud once to check that there are no bizarre names or expressions inserted by the AI.

Five everyday safety rules

Summarizing this entire book in five lines yields the following. These are fundamental habits that reduce risks when using AI for small everyday tasks:

1. AI fabricates answers. Use it for writing and explanations, but verify facts (Chapter 1).
2. Ask specifically as if making a request, and refine your prompt. State circumstances, but exclude personal identity (Chapter 2).
3. Recheck numbers, dates, names, and locations with official institutions. Address health, law, and money with professionals (Chapter 3).
4. Never input red-level information. Resident registration numbers, passwords, verification codes, and account/card numbers (Chapter 4).
5. Hang up on urgent demands for money, and verify with a saved number. Voices and faces are not evidence (Chapter 5).

Going deeper

Procedures are stronger than detection technology

The question "Can't we use AI to detect fakes made with AI?" is frequently asked. Technology to detect deepfakes is being researched, but the technology that creates them and the technology that detects them are in a relationship of chasing and being chased. A method that detected them well today may not work tomorrow. Furthermore, you cannot run a detection program while answering a phone call that comes in the middle of the night.

On the other hand, the procedure of "hanging up, calling back at the saved number, and asking the prearranged question" does not easily break down even if imitation technology improves. This is because even if a scammer perfectly imitates my daughter's voice, it is far more difficult to orchestrate the entire process up to the daughter answering the phone that I dialed directly using the saved number and answering the verification question. However, this procedure is not perfect either. As seen earlier, if a malicious app is installed on my mobile phone, even the calls I make can be intercepted, so in such cases, verify using someone else's phone. That is why this chapter taught "procedures" before "an eye for spotting fakes."

This principle is the same as banks verifying multiple times when transferring large amounts of money. Even if a bank employee knows the customer's face well, large sums of money are verified through established procedures. Human eyes and ears can be deceived, but well-designed procedures are not easily fooled.

When sharing what I created with AI

Now I, too, can create pictures with AI, turn photos of my grandchildren into cartoons, and make celebratory videos. It is an enjoyable thing. However, when sharing what I have created with others, it is good to follow a few practices:

• State that it was made with AI. The labeling obligation under the AI Basic Act applies to business operators that provide AI services. However, if individuals also add a brief note such as "This is an image made with AI," the recipient will not mistake it for a real photograph. This is especially true for photos or videos that are difficult to distinguish from reality.
• Do not play pranks with another person's face or voice. Even a fake video made for fun can cause great hurt to the person involved, and if spread, it can be used for fraud or bullying. As seen in Chapter 4, do not use other people's photos without their permission.
• Narrow the visibility settings for family photos. As in the National Police Agency guidance seen earlier, publicly exposed faces and voices can become material for deepfakes. It is safer to post family photos and videos so that they are visible only among family members.

At work

Ms. Ok-sun Yoon's late-night phone call

Situation. Ms. Ok-sun Yoon, whom we saw in the introduction, was asked for KRW 5,000,000 in a phone call that came in with her daughter's voice.

Spotting the signals. All five signals were present in this phone call: "Right now" (urgency), "You must not tell anyone else" (secrecy), "Settlement money of KRW 5,000,000" (money), and an unknown number (unfamiliar number). While there was no link or app yet, the moment she agreed to send money, they would have provided an account number.

Discarded options.

• Send the money first because it is definitely her daughter's voice. Discarded because a voice is not proof.
• Ask the other party, "Are you really my daughter?" Discarded because the scammer would only say "Yes," and the longer the call lasted, the more they would intimidate her.
• Make a verification call to the "police number" provided by the other party. Discarded because that number could belong to the scammer's accomplices.
• Tell no one in order to resolve it alone. Discarded because the demand to "keep it secret" was itself a signal of a scam.

Decision. Ms. Ok-sun Yoon hung up the phone and called her daughter's saved number. When her daughter answered, she asked the verification question agreed upon during Seollal, and her daughter answered immediately. After confirming that her daughter was safe, Ms. Yoon also notified her son. Together with her son, she reported it to 112 and reported the incoming phone number to 1394. The next day, she posted a message in the family group chat: "I received a phone call like this. Let us remember our promise once again." Her daughter changed the singing videos she had posted publicly so that they were visible only to friends.

Ms. Gyeong-ja Seo's group chat investment video

Situation. Ms. Gyeong-ja Seo, the general secretary of the Neulbom Mountaineering Club featured in Chapter 2, saw a video posted in the member group chat. It was a video in which an economics expert frequently seen on television said, "If you sign up for this product now, you will triple your principal in three months, with principal guaranteed. Recruiting members only until this week." Below the video, a sign-up link was attached. Already, two members had replied, "Thank you for the great information."

Spotting the signals. Ms. Gyeong-ja Seo recalled the signals she had learned in Chapters 3 and 5: "triple the principal" and "principal guaranteed" (guaranteed high returns), "only until this week" (urgency), and a sign-up link (link). The fact that the face in the video was a celebrity was not proof.

Discarded options.

• Trust it because a celebrity spoke directly. Discarded because faces and voices can be imitated with deepfakes.
• It is real because there is no "AI-generated" label on the video. Discarded because the absence of a label does not mean it is real.
• Click the link to check what kind of product it is. Discarded because clicking the link itself could be dangerous.
• Pretend not to notice to avoid spoiling the mood unnecessarily. Discarded because members could lose money.

Decision. Ms. Gyeong-ja Seo searched the economics expert's official channels and the news. There were no reports that the expert had recommended such a product, and there was instead an announcement warning people to beware of investment advertisements impersonating that person. Ms. Seo posted this in the group chat: "Everyone, this video is very likely a fraudulent advertisement imitating a celebrity's face. Please do not click the link, and if anyone has already clicked it or sent money, call your bank immediately." She also called the member who had posted the video separately, explained the situation, and asked them to take down the video. She reported the linked website to 1394. She refined the wording of the group chat announcement by prompting AI to write "so that members do not feel offended, but clearly." Ms. Seo decided the warning content and judgment herself, while the AI assisted with the tone.

Common mistakes

1. "The voice was really my daughter's, so it must be my daughter." Voices and faces can be imitated with AI. Verify through procedures, not by what you hear and see.
2. "The bank's number appeared on the screen, so it must be the bank." Caller IDs can be manipulated. Hang up and dial the official representative number yourself directly.
3. "The other party knows my name and address, so they must be a real institution." Scammers sometimes approach already knowing personal information. The fact that they know it is not proof.
4. "I checked by asking 'Is that really you?' on messenger." The scammer will merely answer "Yes." Verify by calling the saved number.
5. "There is no AI label, so it is a real video." The labeling obligation falls on business operators, and scammers do not comply with it. The absence of a label does not mean it is authentic.
6. "It is an investment recommended by a celebrity, so it is safe." Investment advertisements featuring celebrity faces are frequently used in deepfake scams. "Guaranteed high returns," "principal guaranteed," and "only right now" are suspicious signals.
7. "If money has already been sent, it is too late, so reporting is useless." Applying for a suspension of payments with the financial institution immediately can reduce damages. The faster, the better.
8. "A text summarized or translated by AI is identical to the original." Summarization can alter or omit numbers, and translation can change the meaning. Check dates and amounts against the original text, and back-translate important translations to verify them.

Summary at a glance

Check the official docs

• Financial Supervisory Service Voice Phishing Keeper — What to Do in Case of Voice Phishing Damage: Suspension of payments, measures upon Privacy exposure, deadline for damage relief application.
• Financial Supervisory Service Voice Phishing Keeper — Prevention Tips: Ten prevention tips, including demands for financial information, luring to automated teller machines, caller ID manipulation, and website address verification.
• Financial Supervisory Service Voice Phishing Keeper — Key Contacts in Case of Phishing Damage: 112, 1332, 118, and financial institution customer service numbers.
• Policy Briefing — Report Number '1394' of the Joint Response Center for Telecommunications Financial Fraud (National Police Agency, February 6, 2026) and Six-Month Achievements Since the Launch of the Joint Response Center (National Police Agency, April 2, 2026): Consultation, reporting, and coordinated actions of 1394, 24/7/365 consultation, and tactics impersonating prosecutors.
• Ministry of Government Legislation Easy-to-Find Practical Law Information — Tips for Preventing Electronic Financial Crime Damage: Warning that calls made from a phone with a malicious app installed can be intercepted, and methods for verifying using another phone.
• Policy Briefing — Beware of AI Phishing Imitating Your Family's Voice!: Hanging up and verifying with the person directly, being cautious about posts containing voice recordings (National Police Agency, May 27, 2024).
• Policy Briefing — Beware of Fake Video Financial Scams of 'Child abduction' Using Deepfakes: How to report when it is difficult to hang up, reducing public posts (National Police Agency, November 7, 2024).
• National Law Information Center — Article 31 of the AI Basic Act: Prior notification of AI use, labeling of generated content, and notification and labeling of outputs that are difficult to distinguish from reality.

Self-check

5-1. Write down all five warning signs of a scam contact.

5-2. Write down the "Stop, Hang up, and Verify" procedure in order across its six steps.

5-3. Write in one or two sentences why you must not verify through that call even if the bank's main telephone number appears on the screen of an incoming call.

5-4. Write down two examples that are unsuitable for use as our family verification question, along with the reasons why. (Do not write down the actual question and answer you established.)

5-5. When you see a "video without a label indicating it was made with AI," write one sentence each for what can be known and what cannot be known from the fact that this label is absent.

5-6. When you realize that you have already sent money to a scam account, write down the very first thing to do and two things to do after that.

5-7. When asking an AI to summarize a long notice, write down one thing to do before sending it and one thing to do after receiving it.

Hands-on tasks