Table of Contents
- What is AI
- Danger of AI
- Japanese law on AI
- Revision of the Unfair Competition Prevention Law
- Revision of copyright law
- Overseas Laws Concerning AI
- US case
- Case in Europe
- Two points of contention that are important in AI copyright issues
- Are things created by AI copyrighted?
- Are AI programs copyrightable?
- Ethical Issues to be Discussed in Considering AI Laws
- liability for accidents
- AI’s bias problem
- 5 books on AI law
- 3 papers on the law of AI
- 1, Examination of legal personality of AI in tort law (Ryota Akasaka, 2018)
- 2, Fundamental Ideas Concerning Laws and Regulations for AI (Kaoru Morita, 2017)
- 3, Autonomy of AI and necessity of robot law (Fumio Shimbo)
- summary
What is AI
First, let me explain what AI is.
”AI is an abbreviation for artificial intelligence. Due to the significant improvement in computer performance, computers, which are machines, can now “learn”. That is the core technology of current AI, machine learning.
Danger of AI
Recently, the number of businesses using AI is increasing. The problem is that we cannot completely deny the possibility of AI causing damage to people and other businesses.
If AI causes harm to others, what kind of legal issues should be dealt with? Also, where is the responsibility?
Japanese law on AI
What does Japanese law say about AI? Below are two specific examples.
Revision of the Unfair Competition Prevention Law
The revised Unfair Competition Prevention Act came into force on July 1, 2019.
In the past, it was difficult to prevent unauthorized distribution of items that were not subject to protection under the Patent Act or Copyright Act, or that did not correspond to “trade secrets” under the Unfair Competition Prevention Act.
Therefore, civil measures (the right to demand an injunction, an estimate of the amount of damages, etc.) were stipulated for highly malicious acts such as illegal acquisition or illegal use of certain valuable data.
Revision of copyright law
Under the conventional copyright law, it was possible to use a copyrighted work as long as you performed “information analysis” yourself. However, while the term “information analysis” was thought to include machine learning, including deep learning, the law did not specify it.
This has led to the issue of copyright infringement. Due to the revision of the Copyright Act, the definition of “information analysis” was established, making it legally clear that “information analysis” includes machine learning including deep learning.
Overseas Laws Concerning AI
Next, I would like to introduce some examples from overseas.
US case
The White House announced “PREPARING FOR THE FUTURE OF ARTIFICIAL INTELLIGENCE” in October 2016.
This is a report, not a law, but the application of artificial intelligence for the public good; policies and guidelines for the use of artificial intelligence in the federal government; Economics, artificial intelligence and fairness, safety, governance, security, etc. are discussed.
Case in Europe
The European Commission published the Proposal for a Regulation laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) on 21 April 2021.
It was proposed with the aim of ensuring that AI systems used in the EU market are safe and respect existing laws and regulations regarding fundamental rights and EU values.
Two points of contention that are important in AI copyright issues
Next, I will introduce two points of contention between AI and copyright.
Are things created by AI copyrighted?
When discussing AI and copyrights, the question often raised is whether there is a copyright on AI’s ‘creations’.
if created by humans
Literature, science, art, music, etc. that express the thoughts and feelings of the author are called works, and human creations are recognized as works.
Created by humans using AI
According to the Japanese government, human involvement is required for a work to be recognized as a work, and only creations involving human involvement are considered copyrighted works.
Therefore, in this case, as long as humans use AI as a tool, it can be said that it is a “copyrighted work” because there is human involvement, and it is protected by copyright.
Created by AI
The things created by AI are characteristic and original, so it can be said that they are “creative expressions”.
However, the problem lies in the part of the text of the law that says “thoughts or feelings.” From here, the question arises as to whether AI has “thoughts and emotions”. From this point of view, those created by AI are not recognized as copyrighted works.
Are AI programs copyrightable?
Next, we take up the question of whether the programs that make up AI are copyrightable.
When protected
AI is one of the programs. Therefore, if you obtain a patent for an AI program, it will be protected under the Patent Law.
In this case, if the program is specifically implemented using hardware, it will be protected by software patents.
If not protected
From the above, it can be seen that AI programs that are not specifically realized using hardware are not protected by patents.
So, is the trained model that has finished machine learning protected?
According to Japanese court precedent, in order to be recognized as a program under copyright law, it must be a combination of instructions that cause a computer to function and achieve a certain result. increase.
A trained model is a list of numbers as a result of calculation through machine learning. Based on such a view, it can be said that a trained model is not a program under copyright law, and it is highly likely that it will not be protected under copyright law.
Ethical Issues to be Discussed in Considering AI Laws
Next, we discuss various issues of AI and ethics.
liability for accidents
Who is responsible for the problems caused by AI?
First, it is the users and manufacturers of AI that can be held responsible. It is “tort liability” that AI users may be liable. However, for liability to arise in this case,
| Infringement is intentional or negligentthat damage has occurredThere is a causal relationship between the infringing act and the damage |
You must meet the above requirements. Therefore, if an action by AI causes harm to another person, it is unlikely that the AI user will be held liable for tort due to negligence unless it was intentional.
In addition, the business that manufactured AI may also assume “product liability” as a manufacturer. in this case,
| that there is a defect in the product;damage to othersthat there is a causal relationship between the product defect and the damage; |
Liability arises if the above requirements are met. The difference here from the user’s liability is that the manufacturer’s negligence is not taken into account.
This means that the manufacturer will be held liable whether intentionally or unintentionally. Manufacturers of AI must prove that they could not have foreseen the state of the art of AI at the time it was developed or sold.
AI’s bias problem
The AI bias problem is the error in which certain elements in a dataset are considered more important or representative than others. Biased datasets do not accurately reflect the model’s use case, skewing results and leading to inaccuracies and erroneous analyses.
The problem is that this bias causes AI to make unfair judgments based on gender and race.
For example, an AI system made with image data that has more white people than black people is good at recognizing white people, but it has difficulty recognizing black people.
5 books on AI law
In this chapter, I will introduce five books on the law of AI that I recommend.
1, Introduction to law in the age of AI – an interdisciplinary approach
Each chapter deals with the areas that the study of law must deal with, such as the rapid changes in the insight into the interaction between law and society, and the frequent occurrence of legal problems.
2, Robot/AI and law
It presents knowledge from the study of law, with a view to problems that are currently occurring as well as problems that may occur in the near future.
3. AI law
It also includes new AI legal issues, such as legal issues related to AI and machine learning in financial institutions, and shows the latest practices.
4, IoT /AI laws and strategies
In addition to explaining the legal issues that companies should resolve when using IoT , AI, big data, etc., the prospects for corporate legal strategies toward the realization of an IoT society are written.
5, The legal forefront of AI, IoT , and big data
Based on the latest revisions and guidelines, it clarifies points to keep in mind regarding legal practices for new businesses that utilize the latest technology, protection of intellectual property and data, and ideal contracts.
3 papers on the law of AI
Here, although it is somewhat advanced, I will introduce recommended papers.
1, Examination of legal personality of AI in tort law (Ryota Akasaka, 2018)
Issues that arise with the development of autonomous AI are being considered. The author has been working as an associate professor at the Center for Co-Creation of Social Technology, Osaka University since April 2020.
2, Fundamental Ideas Concerning Laws and Regulations for AI (Kaoru Morita, 2017)
It is written about laws and regulations as AI develops. The author has been a professor at Tohoku University since 2015.
3, Autonomy of AI and necessity of robot law (Fumio Shimbo)
The need for robot law is stated from the challenges of AI. He is a professor at the Faculty of Policy Management, Keio University.
summary
So far, we have discussed AI and legal issues. How was it?
Do you know what the legal liability of AI looks like and how AI is protected by law?
