Privacy, Copyright and Ethics — Using AI Responsibly

1.Personal information: once you enter it, it leaves your hands

What you type into a conversational AI is sent to the service provider's servers and processed there. Depending on the service and your settings, conversation history may be stored, reviewed by people to improve quality or used for training. How it is handled differs from service to service, and policies change. So it is safest to decide whether to enter something based on one question: "Would it be all right if other people saw this information?"

In particular, do not enter national ID numbers, passport or driver's license numbers, bank account or card numbers, passwords and verification codes, health information, or exact addresses and contact details. This applies not only to your own information but also to that of your family, customers and coworkers. Entering other people's personal information without their consent can violate their rights, and at work it can lead to legal problems.

• Don't enter: national ID numbers, bank account and card numbers, passwords, verification codes, health information, exact addresses and contact details
• Enter in altered form: names as A and B, your company as "a manufacturer," and only as many numbers as needed
• For company material: first check which services and rules your company allows
• Check your settings: look at the service's settings and policies to see whether you can turn off conversation history and use for training

• Wrong: Write a reply to the complaint email from customer John Doe (010-1234-5678, ○○ District, Seoul).Right: Customer A complained that "the delivery was a week late." Write a polite reply that apologizes and says we will send it again by next Tuesday.You don't need the name and phone number to write the reply. Keep only the information the task really needs, and delete or replace the rest.

2.Copyright: being able to make something doesn't mean you may use it

AI creates new output from the patterns of the countless texts and images it learned from. This raises two questions. One is whether AI output is too similar to someone else's work; the other is who holds what rights to AI output. On both questions, laws and rulings differ from country to country, and debates and lawsuits are still ongoing.

So it is wise to set conservative principles you can actually follow. Avoid requests to reproduce, as if copying, the sentences of a particular work or a particular character, and requests to output song lyrics or long passages of a book word for word. When you use AI output commercially or publish it, check whether the terms of the service you use allow that purpose.

You should also know that output largely made by AI may not be protected as a work created by a person. In many countries, copyright is seen as protecting human creative expression. For important creative work, keep a record of the parts you planned and revised yourself, and consult an expert if needed.

3.Bias: skewed data becomes skewed answers

As we saw in Lesson 2, learning-based AI learns from data. If the data contains society's stereotypes or skews against particular groups, the AI learns them too. For example, when asked to describe occupations, it may portray gender or age only one way, or give less accurate answers about certain regions or languages.

Bias is hard to spot, so you need to be all the more careful. If you use AI for decisions that affect people, such as hiring, lending or evaluations, you must check separately that the results are not skewed against particular groups. In everyday life, too, "the AI said so" is no proof of fairness.

4.Responsibility: whoever sends it out is responsible

AI is a tool, and it is a person who decides to use the result. If a report written by AI has wrong numbers, the person who submitted the report is responsible. If text produced by AI defames someone or infringes copyright, the person who posted it can be held responsible. "The AI wrote it that way" is no excuse.

So for anything that goes out under your name, check three things: Are the facts and numbers correct (Lesson 7)? Does it infringe anyone else's rights? Could it harm or mislead the people who read it? Also judge, according to the situation, whether to disclose that you used AI. If it is something readers need to know in order to make their own judgment, disclosing it is how you keep their trust.

You also need to be on guard against the misuse of AI. Fake videos and audio that imitate a real person's voice or face, and convincing scam text messages, have become easier to make thanks to AI. If someone suddenly contacts you asking for money or personal information, always verify through a different channel, even if the voice or writing style seems familiar.

5.A checklist for before and after using AI

The checklist looks long, but once it becomes a habit it takes only a few seconds. At first, apply it only to important tasks, and once you are used to it, extend it to everyday use.

6.Applying the checklist to a real situation

A checklist only sticks once you have actually applied it. Below is an example that follows one common, everyday situation in order: before entering, after getting the result and before sending it out. The situation is made up for illustration.