The intersection of technology and law comprises topics with vast implications, from the content of patents and copyrighted material to the practice of law itself. Just one aspect of technology that can create legal issues in these areas is the use of artificial intelligence (AI), particularly undetectable AI. The rapidly advancing ability of AI to generate content that cannot be traced back to its source or distinguished from the work of a human will have irreversible impacts on the practice of law, and it's important for the legal minds of our world to be well aware of it. The potential misuse of AI-generated content is something practicing lawyers and online JD students will need to wrap their heads around for years to come.
This article will explore the complex world of AI and AI detection before exploring legal issues of copyright infringement and trademarks.
Detecting AI
Text generated by large language models like Chat GPT already often matches the quality and characteristics of human writing to the point that it has ostensibly passed the infamous Turing Test. The Turing Test is a benchmark for computer performance set by computer scientist Alan Turing. It essentially determines whether a machine's ability to show intelligence is indistinguishable from that of a human.
In the months since the study was published, more than a dozen AI models have been launched by as many companies that either match or exceed the performance of the original GPT-4, and most are available for free use.
It doesn't stop there. OpenAI, the company behind GPT-4, recently revealed a product called Sora, an AI platform capable of generating photorealistic video. Midjourney, another AI tool, can now generate photorealistic images that can be nearly impossible to discern from photos.
It's indisputable that AI is working its way into many parts of our lives, including the media and content we consume. But that alone doesn't mean that AI-generated content can't be detected. Universities, publishers, and technology companies have raced to find solutions to the host of problems inherent in the ensuing wave of AI-generated content and have been able to isolate patterns in the writing styles of large language models, which have been used to produce tools to make doing so accessible. Even AI-generated video can be reliably detected with over 90% accuracy. Other tools have been used to detect AI-generated photos. Unfortunately, as in any part of the technology industry, it seems that bad actors are innovating just as rapidly and effectively as those seeking to retain the credibility and relevance of our existing institutions, and text generators designed to disguise AI-generated text from these AI detectors are springing up.
Intellectual Property
AI-generated content will, without a doubt, have a lasting impact on the framing of intellectual property. At the time of writing, copyright law currently dictates that authorship protections are reserved for content produced by humans. However, with AI-generated content becoming more and more difficult to separate from what we as humans generate with our own hands and minds, this historically vital piece of legislation may become unenforceable. While detection can potentially be used in the process of validating or enforcing some copyrights, not all content that would be covered by copyright, such as slogans and other short-form content, can feasibly be discerned as AI-generated or not.
Trademarks are an even trickier subject. While there is no explicit law or provision that limits a person's ability to trademark content generated by AI, or especially content that contains AI-generated elements, the eligibility of AI content for trademarks may come into question in the future. Intellectual property laws worldwide are currently being debated, revised, and reinforced to wrangle with the influx of content created using generative AI. In theory, since generative AI is trained on the works of others, it may be decided in some jurisdictions that AI-generated content is accepted as a derivative product of other content that is often copyrighted or trademarked. But this would remain contingent on our ability to detect AI-generated content in a way that would be admissible as evidence in a court of law.
Regulation
How can these risks be mitigated in a measured, enforceable fashion that encourages accountability without putting the benefits of the technology at risk? At present, it seems that, given the indistinguishability of some AI-generated content from that created by humans, it may currently be impossible to regulate its use in order to determine the validity of intellectual property or to hold its users accountable for actions that are, or may become, breaches of the law.
This problem might be completely insoluble through current legal means. The one technology that shows some promise of helping aid the regulation of generative AI content is the idea of watermarks. The World Intellectual Property Organization, or WIPO, has proposed the use of blockchain watermarks embedded in digital files as a way to track digital assets and potentially even incentivize informants. While blockchain is just one specific solution, with limited implications in certain forms of media, the idea of watermarking has broader potential. It's been suggested that AI-generating visual content could be required to embed visual watermarks into their images as a way of ensuring that they are either impossible or at least very difficult to use without detection.
The degree to which any solutions to gray areas in AI usage, whether ethical, legal, or otherwise, can be implemented effectively and consistently remains unclear. With multiple state-of-the-art models being made publicly available by big tech companies like Meta and Mistral, the notion of any regulation that requires compliance from generative AI companies may be incredibly difficult to enforce or monitor. Only time will tell if solutions will be built that have a more universal ability to detect generative AI content; in the meantime, the prospect of regulatory efficacy remains uncertain. Like many things in the science and technology worlds, AI is a rapidly evolving area, becoming more powerful and more complex by the day—it's difficult to know for sure what the future holds.
