The use of interactive AI "ChatGPT" and image generation AI is rapidly expanding.
Guidelines with precautions when using it are published by the Japan Deep Learning Association.
While anyone can easily use the generated AI, the risk of information leakage and infringement of rights such as copyrights has been pointed out.
Generative AI, what should you watch out for?
At a minimum, we will look at the precautions that you need to know with reference to the guidelines.
(The guidelines will be updated from time to time, so please check the latest information)
"Generative AI" will be introduced not only by individuals but also by companies and organizations
"Generative AI" is "artificial intelligence" that has the ability to generate sentences, images, music, etc. based on a large amount of data trained in advance.
Even if you do not have specialized knowledge, you can generate sentences, program codes, images, etc. just by inputting simple "instructions" with text.
It is considered to be applied in various fields such as business, marketing, education, entertainment, such as summarizing meeting minutes, translating, and generating ideas, and has begun to be introduced by many companies and organizations other than individuals.
There are two main points to note.
There are two main points to note when using generative AI.
(1) Copyright Be aware of the possibility of infringement of trademark rights and other rights.
The guidelines list "copyrights," "trademark rights," "design rights," and "publicity rights" as "special care should be taken when entering the following types of data from the viewpoint of the necessity of processing intellectual property rights and compliance with laws and regulations."
Regarding copyright, while stating that "simply inputting another person's work into AI does not constitute copyright infringement," it is stated that "if the generated data is the same or similar to the entered data or existing data (copyrighted work), the use of the product may constitute copyright infringement of the work."
The same applies to trademark rights and design rights, which are "Commercial use of trademarks or designs that are identical or similar to the registered trademarks or designs of others intentionally or accidentally generated is trademark infringement or design infringement."
So, what kind of work should people be aware of?
Particular attention should be paid to the so-called "creators".
"AI artists" who use AI to create images such as illustrations and designs.
"Copywriters" and "designers" who come up with catch phrases, logos, designs, etc.
Also, be careful about the "poster" of the video.
Overseas, a song that reproduces and imitates the voice of a popular artist with AI went viral on SNS and became a hot topic, but the management company of the artist's song asked for its removal from the streaming service as "copyright infringement", and it was subsequently deleted.
In addition, in the case of image generation AI, there is a movement to file a class action lawsuit against the AI operating company in the United States, claiming that "my work was used for AI learning without permission and a similar work was created."
In Japan, when a group made up of actors and musicians conducted a questionnaire survey, specific rights violations such as "the drawing style was plagiarized," "the published manga was used as data for AI to learn without permission," and "my voice was sold as a model for AI processing without permission" as a specific infringement of rights.
(2) Information entered into the generated AI may be leaked to others
Care must also be taken when inputting information into the AI.
The guidelines call for people not to enter so-called "highly confidential information" such as "personal information, confidential information, and confidential information."
The data entered by the user may be used to train the AI model.
It has been pointed out that if highly confidential information is entered, the contents of the information may be leaked to companies that provide generative AI services and other users.
The following people who have many opportunities to handle personal and confidential information should be especially careful.
"Medical professionals" and "public officials" who handle personal information, as well as "educators" and "judicial officials".
"Sales and contract personnel" and "financial and technical development personnel" who are involved in the secrets of companies and organizations.
If you use "the person in charge of meeting minutes" because it is convenient for summarizing, the contents of the meeting may be leaked.
In addition, industries that handle confidential information include those involved in "media" and "security".
In the research of a domestic security company, when they instructed the generated AI to use a special method called "jailbreak" and requested the disclosure of another person's email address, they disclosed an address that would never be disclosed as "privacy-related".
Leaked information may be misused for crimes, etc.
It has also been pointed out that the leaked information may be further misused for cybercrime and other purposes.
Overseas, there have been a series of cases where employees of major companies have entered confidential company information into the generation AI, and according to a questionnaire conducted by a security company, it was found that about 7.5% of people had entered company data.
Some of them were confidential internal information, followed by source code and customer data.
When using generative AI, it is also important to determine in advance the rules to what extent it can be used within a company or organization.
In the first place, the content generated by AI may be false
So far, we have looked at "precautions when using", but there are some things that you must always check when "using" the generated one.
"AI-generated content may contain falsehoods,"
the guidelines point out.
"The principle of large-scale language models is to create plausible sentences by outputting the word that is most likely to be used after a certain word."
For this reason, the generated content is not always "correct".
The guidelines state that it is important to "know the limits of generative AI, do not blindly trust the contents of the product, and always confirm the basis and support yourself."
Ministry of Education, Culture, Sports, Science and Technology (MEXT) is considering guidelines for educational settings
In addition, with regard to use in educational settings, where there are concerns about the effects of plagiarism and loss of thinking ability, the Ministry of Education, Culture, Sports, Science and Technology (MEXT) aims to create guidelines by around this summer on the appropriate age for use, when it should be prohibited, ideas for classes to learn how to use it and how to use it, and how to reduce the burden on teachers on school duties.
"AI Strategy Team" to discuss with relevant ministries and agencies
In addition, the AI Strategy Team, which discusses AI regulations and how to utilize AI among relevant ministries and agencies, will also discuss the revision of guidelines for corporate users of the Ministry of Internal Affairs and Communications and development companies of the Ministry of Economy, Trade and Industry (METI) based on the spread of generative AI.
Every day, various users raise information about new uses and discoveries, as well as points of concern, regarding generative AI.
It is desirable for companies and organizations to incorporate the latest information, proceed with the creation of rules as appropriate, and use it carefully while referring to various guidelines.
Japan Deep Learning Association will update the published guidelines from time to time.