Supreme Court of India Implements AI for Transcribing Court Proceedings, Raises Concerns of High-Risk AI Under Proposed EU Act
The Supreme Court of India has recently decided to incorporate artificial intelligence (AI) technology in the transcription of live court proceedings. In a bid to enhance efficiency and accuracy, the court called for sealed bids from experienced firms to design, develop, and implement AI solutions for transcribing arguments and court proceedings.
However, this move has raised concerns in light of the proposed European Union (EU) AI Act, which categorizes this application as high-risk AI. The EU aims to ensure the safety and adherence to fundamental rights and values of AI systems within its member states.
Kshitiz Verma, an Assistant Professor at JK Lakshmipat University, commends the EU AI Act, stating that it provides comprehensive guidelines for the responsible use of AI. Verma emphasizes the need for caution, referencing a previous incident involving court employees who mistakenly omitted a crucial word in the Anil Ambani case. This error resulted in their termination and highlighted the significant impact even minor mistakes can have on legal proceedings.
While significant progress has been made in the development of speech recognition technology, it still lags behind advancements in vision and natural language processing. The accuracy of AI transcription models remains imperfect, with error rates around 12 percent. Furthermore, the diverse accents in India pose a unique challenge, making it even more difficult for AI to accurately recognize words in a noisy and chaotic courtroom environment.
Improving the accuracy of transcriptions necessitates a wide range of training data, especially considering the multitude of Indian accents. Similarly, the success of natural language processing and machine translation varies across languages. Languages like Hindi, commonly used in Indian courts, fall into the lower resource class, resulting in smaller datasets and limitations in translation capabilities.
To mitigate the risks associated with AI systems, Verma emphasizes the need for increased human oversight. Those overseeing AI must possess sufficient AI literacy and authority to conduct thorough investigations when necessary. This oversight ensures that risks to health, safety, rights, and the rule of law are minimized, preventing automation bias and unquestioning acceptance of AI outputs.
Verma stresses that any AI system violating the fundamental rights and safety of citizens should be deemed high-risk and subject to extensive conformity and precautionary measures.
While the implementation of AI in transcribing court proceedings offers potential benefits in terms of efficiency and accuracy, it also raises concerns about the risks associated with such high-risk AI systems. Striking a balance between technological advancements and maintaining the integrity of legal proceedings is crucial. It is necessary to proceed with caution, ensuring that human oversight and adherence to guidelines are prioritized to guarantee the fairness and accuracy of court transcription and translation processes.
Overall, the Supreme Court of India’s decision to implement AI for transcribing court proceedings highlights the increasing role of AI in various sectors, including the legal field. However, it also underscores the need for comprehensive legislation, like the proposed EU AI Act, to govern the responsible use of AI and mitigate potential risks.