Generative Artificial Intelligence has moved from research labs to mainstream adoption, with tools such as ChatGPT, MidJourney, Stable Diffusion, and GitHub Copilot driving product development, creative industries, software engineering, and scientific research. That rapid diffusion tests traditional intellectual property frameworks built around human authorship and inventorship, creating new risks for creators, companies, and legal practitioners as the line between human and machine contribution blurs.
The article outlines three core IP challenges. First, authorship and inventorship rules presume a natural person, a point reinforced by the U.S. Copyright Office in 2023 and by patent office decisions in the U.S., UK, and EU that rejected AI as an inventor in DABUS filings. Second, training data rights raise questions about fair use, consent, and licensing because generative models are often trained on copyrighted material. Third, co-creation scenarios require clear allocation of ownership and documentation of human input to enable protection and commercialization. The piece lists jurisdictions including Australia, Canada, Germany, Israel, New Zealand, South Africa, South Korea, Switzerland, the UK, and the U.S., noting that while some systems will recognize patentable subject matter linked to AI, a natural person typically must be named as inventor and that South Africa has permitted filings with AI-related inventorship. It also notes China has shown greater openness and that the EU is advancing regulation under the AI Act with an emphasis on transparency and accountability.
On copyright, the article highlights two tensions: the human authorship requirement that limits protection for purely machine-generated works and the derivative works risk when outputs mirror copyrighted training material. High-profile disputes such as Getty Images v. Stability AI and Andersen v. Stability AI/MidJourney illustrate these issues. For patents, practical hurdles include inventorship standards, enablement and disclosure problems posed by black-box models, and determinations about patentable subject matter, with a recommended focus on patenting enabling tools and ensuring human involvement.
The article also describes how Artificial Intelligence can enhance IP practice, listing use cases for prior art search including natural language processing, semantic search engines, automated patent clustering, cross-lingual search, generative summarization, and visualization tools. It cautions that regulators expect human oversight and accountability when these tools are used. For protecting data and trade secrets, the authors recommend licensing or creating proprietary datasets, combining patents and trade secrets where appropriate, and developing output licenses and cross-licensing arrangements to monetize innovation while managing rights across supply chains.
To build a practical IP roadmap the piece advises auditing Artificial Intelligence use cases, documenting meaningful human involvement, securing data rights, mixing protection mechanisms, developing internal policies on employee use and authorship, and monitoring evolving case law and regulation. Organizations that blend human creativity with Artificial Intelligence capabilities while maintaining robust documentation and region-specific strategies will be best positioned to safeguard innovation and capture commercial value in this fragmented legal environment.
