Most of the industry leaders as well as AI Ethics advocates would agree with Mira Murty (CTO, OpenAI) when she stated:
“AI systems are becoming a part of everyday life. The key is to ensure that these machines are aligned with human intentions and values.”
Legal Implications of CHATGPT
The legal implications of CHATGPT in the global as well as the national perspective include AI regulations, Data protection laws and copyright consideration. This is not an exhaustive list since it is still an evolving topic, both in the tech as well as the legal world.
In this article, we will delve into the legal implications of using Large Language Models (LLMs) or other systems based on Neural Networks in the global and national perspectives. To corroborate whether the aforementioned machines are well aligned with the general human intentions as well as the moral code of conduct, let us discuss the ways in which the OpenAI tools operate.
The OpenAI Charter which is a set of principles used by OpenAI for the execution of its operations, aims to provide safe and beneficial Artificial General Intelligence (AGI) for everyone, keeping the best interest of humanity in mind. It seeks to create a global community with constructive competition among the contemporaries, sharing safety policy and standards research.
OpenAI has a simple and straightforward Usage Policy where users are expected to follow a deliberate system of guidelines applicable within law. These include,
not using the services to harm oneself or the others,
not remodeling the outputs of the services to harass others and respecting the company safeguards.
Building with OpenAI should comprise best practices without compromising the safety of others. It states that it is not prudent for an individual to make high-stake decisions such as financial or medical, solely based on the outputs of ChatGPT. It further states that one should not indulge in activities facilitating gambling, hampering democracy, spreading misinformation as well as building tools inappropriate for minors.
The “EU AI Act” of May, 2024 stands as the cornerstone of the global perspective around the law on AI since it is the first comprehensive legal AI framework, which makes technology more human-centric and regulates AI. The EU AI Act’s definition of AI is the centerpiece of all the laws formed around it and states that AI is “A machine based system designed to operate at varying levels of autonomy”.
Mitigation of risks related to AI are handled in this Act by ranking the AI systems at low, medium and high risk. For example,
AI systems for equity markets and banking are categorized as high risk infrastructures, and the presence of at least one human supervisor is mandated. It is also open to scrutiny and questioning from the general public.
The “EU AI Act” specifies penalties for the non-compliance of its components:
Type of Non-Compliance | Penalty |
Lying to Regulators | 7.5 Million Euros |
Breaching Provisions | 15 Million Euros |
Developing or Deploying banned ai tools | 35 Million Euros |
ChatGPT stores the chats of the users as training material for future models. The users can delete these chats manually or ask OpenAI to do it for them to prevent them from storing sensitive data. Although the user consents for the aforementioned processes, regulatory and legal discrepancies are always inevitable. To tackle these, OpenAI abides by certain data protection and regulation acts viz:
General Data Protection Regulation
California Bot Disclosure Law
The California Bot Disclosure Law mandates businesses using generative AI in consumer interactions must clearly specify to the consumer that they are conversing with an AI model.
Copyright Infringement is yet another aspect of Intellectual Property Law that AI models like ChatGPT need to take into account. Since the model trains and provides output from various sources on the internet, there is a high possibility that usage of sources may infringe the copyrights of a third party.
The US Federal Trade Commission has urged for the usage of “Algorithmic Disgorgement”, which is a legal remedy that demands companies to retract AI models or data that were obtained or used illegally.
Unintended plagiarism is another concern related to copyright which might emerge due to obscurity around ownership of the output provided by ChatGPT. OpenAI states in its “Terms of Use” section that the ownership of the output belongs to the user, but the outputs provided by ChatGPT are not always unique, since the source is common. This might lead to commercial ownership disputes. ChatGPT does not provide accurate citations either. Thus, it is advised to not write the content of the output verbatim but rather use it as a source of inspiration.
In contrast to the European Union, India does not have a comprehensive legislative framework around AI. The Government directed its think tank the “NITI Aayog” to formulate a national strategy around AI. But a strategy in itself is not enforceable by law. The Ministry of Electronics and Information Technology (MEITY) has formulated an advisory on the applications of AI in India.
Since it is just an advisory, companies like OpenAI are encouraged to follow them when operating in India, but they aren’t enforceable by law either. It advises against the propagation of unlawful content, use of systems or structures with inherent bias, and the use of under-trained or unreliable models. It encourages the companies to ensure that the user is explicitly informed about the possible flaws of the output and unreliability's, if any. This also includes making users aware of the terms of services, and possible consequences of dealing with unlawful components of the infrastructure under the applicable law.
Conclusion
It is evident that organizations deploying AI tools have a formal charter and Usage
Policy in place to uphold the privacy, integrity, and interests of users and other stakeholders.
With the European Union pioneering the development of enforceable laws around AI, countries like India have begun formulating legal frameworks and advisories to regulate AI usage.
However, the lack of comprehensive legislation in many parts of the world, where AI is
increasingly pervasive, creates significant gaps that could lead to serious consequences. To
address these issues, governments, organizations, and developers should collaborate to establish an infrastructure that ensures these technologies adhere to data protection laws and align with human values.
Frequently Asked Questions
Q1. Will my data be deleted if I delete my ChatGPT account?
A1. OpenAI will delete your data within 30 days of your account deletion; it may, however,
retain a certain amount of data for a relatively longer period of time as and when permitted by Law.
Q2. Can I use the output of ChatGPT for my personal or commercial use without any legal
implications?
A2. Yes, as per OpenAI’s Terms and Content Policy, you can use the content for personal or
commercial purposes without legal implications. However, there might be a risk of shared
ownership of content. Additionally, ensure that the generated content complies with local laws and does not cause harm to anyone.
Q3. Is it safe to use sensitive data from my company as a prompt for ChatGPT?
A3. The prompts by users are visible to OpenAI. Although it does not automatically make
the data available for others to query, certain identifiers and parameters can make it available. A paper published by researchers from Apple demonstrated how memorization of data used in training sets can jeopardize the integrity of data through exposed vulnerabilities.
Q4. Is ChatGPT trained entirely on no-copyright sets?
A4. As stated by OpenAI itself in its written evidence for the House of Lords, UK, “it
would be impossible to train today’s leading AI models without using copyrighted materials”. Therefore, it is safe to say that a model which extracts its data by scaling the vast expanses of the internet might include copyrighted material as well.
Q5. Can I use ChatGPT to write my exams?
A5. Assuming the use of tools like ChatGPT is permitted in the examination, a lack of due
diligence when using ChatGPT could expose you to occasional misinformation in the outputs, potentially impacting your grades. It is advisable to study from reliable sources and avoid using ChatGPT outputs verbatim to achieve good marks.
This Article is written by Suvigya Tiwari, a first year law student at Narsee Monjee Institute of Management Studies.
Follow our legal journey on our social media and keep learning.
Insightful,
amazing job👍 n Its impressive to see such a thorough analysis.
this was insightful suvi!
Fabulous...well done..