IP protection in the ChatGPT era

By Wang Yan and Li Yue, Han Kun Law Offices
Copy link

ChatGPT and other artificial intelligence (AI) tools are demonstrating admirable comprehension and conversational skills, enabling them to solve various generic issues in a growing number of areas. Although the technology is not yet self-aware, Michael Kosinski, a professor at Stanford University, concluded after thousands of tests that ChatGPT is able to solve Theory of Mind tasks, reigniting debates on whether AI can be considered a “legal person”.

This article focuses on the protection of AI’s “intellectual achievements”, and how they risk infringing others’ IP rights.


Wang Yan, Han Kun Law Offices
Wang Yan
Han Kun Law Offices
Tel: +86 21 6080 0200
E-mail: yan.wang@hankunlaw.com

Is AI-generated content (AIGC) subject to IP protection? In terms of copyright, the “copyrightability” of AIGC depends on whether the generated content reflects the creator’s originality of expression.

On the one hand, this means that content directly generated by AI is not copyrightable. The Beijing Internet Court once found that the differences between the AI work in question and existing works were only due to data and software choices, or types of graphics. Therefore, the AI work was not copyrightable as it failed to demonstrate the creator’s originality of expression. The US Copyright Office has made similar judgments.

On the other hand, AIGC that reflects the creator’s originality is copyrightable. The Shenzhen Nanshan District People’s Court once found that an article generated by a writing software was copyrightable, as it reflected the originality because of the user’s arrangements and choices. This made it a work subject to the protection of copyright law.

Can AI become an IP rights holder? This question remains controversial but, so far, most jurisdictions have taken a negative stance. Stephen Thaler, the creator of the AI system DABUS, has been filing patent applications around the globe since 2019 that designate DABUS as the “inventor”. This has led to a series of disputes that last to this day.

The inventor status of DABUS has been denied in countries including the US, UK, Australia and New Zealand. Only South Africa registered it as a lawful inventor. Meanwhile, the Registrar of Copyrights has recognised an AI painting app as a co-author in India. AI’s eligibility as an IP rights holder will depend on the dynamic interactions between technology and law, something worthy of our continued attention.


Li Yue, Han Kun Law Offices
Li Yue
Han Kun Law Offices
Tel: +86 10 8516 4204
E-mail: yue.li@hankunlaw.com

Currently, AI mainly interacts with text and graphics, which leads to infringement risks that mainly concern copyright and trade secrets.

Risk of copyright infringement

For developers of AI, the materials used in model training might infringe on others’ legal rights. For example, several artists in the US recently filed a class action at the US District Court of the Northern District of California, accusing Stability AI, Midjourney, and DeviantArt, all developers of various AI art generators, of copying copyrighted images without permission. OpenAI, the developer of ChatGPT, was likewise accused in a class action of scraping copyrighted source code without permission to train AI systems.

Risk of trade secret infringement

If users input trade secrets into an AI tool, the AI tool will utilise such information as training material and, in turn, divulge these secrets in the output to other users, causing leakage of the confidential information. Several companies in the financial sector, as well as other sectors concerning sensitive information, have established preventive measures in this regard.


AI cannot be subject to liabilities under the existing legal framework. Even under the Grishin Law of Russia, which vests AI with legal personality characteristics, the liability lies with the AI’s owner or possessor. Likewise, the European Parliament Resolution on Civil Law Rules on Robotics provides that humans are liable for AI infringements. However, there is no consensus on how exactly the liability should be assigned, with opinions largely divided into the following three groups:

    1. The user is liable as if AI were a tool. As all AI activities are conducted under human instructions, the user of the tool should be held liable for any infringement. This is reflected in the legislative explorations by the Shenzhen Standing Committee of the National People’s Congress with respect to autonomous driving, under which the owner or supervisor of the vehicle shall be liable for torts.
    2. The manufacturer is liable as if AI were a product. Under the Product Quality Law, the manufacturer and seller are held liable for product liabilities. There is a relevant precedent where the Supreme People’s Court found AI software had committed infringements and held the developer of the software (i.e. manufacturer of the product) liable.
    3. The supervisor is liable as if AI were an animal. In this school of thought, AI is considered similar to domesticated animals, which means the supervisor should be held liable for infringement liabilities. For example, any infringement by ChatGPT would be borne by its developer, OpenAI. However, the infringement liabilities of other software with ChatGPT embedded in would be borne by their own supervisors.

ChatGPT is a monumental leap in technology and the existing legal framework is feeling the impact. Legislators and policymakers across the world have responded to the emergence of AI tools to varying extents.

China is exploring regulatory measures, while the EU has added a “general purpose AI” clause into its AI Act, and the US has yet to promulgate any regulatory provisions, with the federal focus lingering on promoting industries.

Both developers and users must take great caution against the infringement risks posed by AI and ensure that their development and usage remain lawful and in compliance with regulations at all times.

Wang Yan is a partner at Han Kun Law Offices. He can be contacted by phone at +86 21 6080 0200 or by e-mail at yan.wang@hankunlaw.com
Li Yue is an associate at Han Kun Law Offices. She can be contacted by phone at +86 10 8516 4204 or by e-mail at yue.li@hankunlaw.com

Back to INTA2023 page for more IP articles

Copy link