Former Bombay High Court Judge Justice Gautam Patel highlighted the need for legislation to regulate ‘Deepfakes’ and flagged the challenges that may arise in addressing the issues with them.
Speaking at LiveLaw’s 5th Shamnad Basheer Memorial Lecture On ‘AI, Trademarks, And Copyright: Emerging Issues‘, Justice Patel said, “New legislations and building awareness about detecting and recognising deep fakes are going to be challenges that lie ahead”.
He then explained the recent trend of ‘deepfakes’, where AI-generated fake imageries dupe the viewer as original images.
“The issue is grave when one considers the generative adversarial neural networks – what happens here is that one machine creates a fake and passes to another machine which detects the flaw in the fake and reports back, thus the adversarial process. The first machine goes to correct the flaws using its generative power and bounces back to the test machine, this adversarial process, not very different from what one sees in law Courts continues in this manner until the final result is a nearly perfect rendition but none the less a fake video, music etc”
“This problem is exacerbated when deepfakes driven by AI are so good that it is impossible for the brand owner, let alone the judge to tell one from the other. In a real world scenario consider what you would have to bring to court to substantiate a case against an indistinguishable AI deepfake. The very prospect should give the IP Bar sleepless nights,” Justice Patel said explaining how such trends can affect registered trademarks and established brand reputations.
Copyright Law And AI : A Zone Of Twilight
Justice Patel, while referring to the recent suit filed by news media ANI against ChatGPT, in the Delhi High Court, addressed a relevant question that seems to be the topic of discussion in the IP field lately- “Who owns the copyright in generated works of art, music, literature created by AI? Is it the user who prompts the AI or the AI itself?”
Notably, Asian News International (ANI) has filed a copyright infringement suit before the Delhi High Court against OpenAI Inc, which founded ChatGPT, alleging unauthorised use of its original news content.
He explained that AI may lack the human work required to get copyright protection which is the ‘sweat of the brow’ principle. The principle mandates that even though a work may not be original, it can be copyrighted if the author has put considerable efforts and skills into making it.
Justice Patel illustrated how AI should not be confused with a tool in helping manifest the creative human work, as it was a creator in itself. He referred to how in 2016 a group of museums and researchers in the Netherlands unveiled a new art piece generated by a computer after analysing 1000s of artworks of the 17th-century Dutch Artist Rembrandt.
Another example was a short novel written by a Japanese Computer Programme in 2016 that reached the 2nd round of the national literary prize. He also mentioned Google’s AI company DeepMind which created a software that can generate music by listening to a recording.
“The point is that AI is no longer just a tool like a paintbrush and easel, things used in the creative process to generate original works that ‘sweat of the brow’ qualify for copyright protection”
“The newest iteration of AI, the algorithm is no tool, it is a creator, and it is bringing into existence a work of and on its own. It is actually making decisions.”
Justice Patel went on to share an experiment he did with ChatGPT prompts, where he prompted the AI to give opening lines sounding like the author Leo Tolstoy. He added, that while the AI gave a generative response, his role in the creation was, as he said, “to create on my own, arguably, entirely originally, the keywords or phrases that generated this result.”
However, it posed another issue which is whether would there be a copyright in his own unique arrangement of keywords and phrases?
He highlighted that this aspect would similarly apply to AI-generated images as well. Using any of the popular AI image generation tools online – one can create seemingly original works based on a unique arrangement of keywords and phrases. In such scenarios, where lies the copyright, Justice Patel wondered?
Terming the present technological advancement a ‘Twilight zone in copyright law’, he emphasized the upcoming challenges for the Court to understand where a line would be drawn to determine infringements.
“This is a twilight zone in copyright law. How on earth would you begin to bring infringement or passing off action or both in a situation like this?”
When it comes to AI images, the challenge is rather doubled as now one can use results generated by AI to feed as prompts to generate AI images, making it a double AI work.
“How much of the original is too much for AI purposes? Remember that in AI image creation we have a sort of double battle – uniquely composed texts are used to trigger an AI-composed image”
Understanding The Potential Challenges In Trademark Vis-a-Vis Artificial Intelligence
Justice Patel elaborated on the recent ‘Pattern Recognition Technique’ used by AI. He said that at the heart of pattern recognition are the computer algorithms designed to analyse and interpret data. Data inputs can be words, texts,inputs or files. Pattern recognition is broader compared to computer vision which focuses on image recognition. He explained that patterns include various trends in the form of repeated data eg: a finger print, a handwritten cursive work, a human face etc.
He further added that the aim of the technique is to replicate the decision-making process of humans which is based on the recognition of patterns for instance – the next move in a chess game is based on the boards current pattern.
Justice Patel analysed that the generative power of AI can put the registered TradeMarks at a risk of losing their uniqueness considering the Pattern Recognition Technology the AI uses to create something new out of the already existing data.
“Because AI analyses vast data sets of language and usage to identify trends and patterns, trademarks are at constant risk of becoming generic quickly reducing their protectable status. Allied to this is the concept of distinctiveness in trademark. As we know that a trademark’s distinctiveness is crucial to its protection, AI can easily generate new and seemingly unique materials capable of trademark protection. Names, logos etc”
The other emerging question of law, according to Justice Patel was “Whether an AI-generated mark is sufficiently distinctive to merit trademark protection.”
Justice Patel suggested how AI could instead be used to the benefit of high-value brands for countering the Trademark (TM) infringements that occur online :
“Can we not use AI to monitor online or digital markets and exchanges, social media and alike to track TM infringement? I believe we could and at least a large number of high-value brand owners need to reposition their in-house capabilities to harness the pattern recognition strength of AI not to create things, but to do these things instead (1) to detect, track and monitor deepfakes; (2) to constantly generate propriety variance on the brand name and logo”
Justice Patel concluded his address by leaving the audience with an explorative thought- “to figure out is this entire address of my evening, original or wholly AI generated?”
The video of the event can be accessed here.
Related News :