Non-Disclosure of data in digital court system : Does tech make it easier to hide information?

/
Articles and Blogposts /
Non-Disclosure of data in digital court system : Does tech make it easier to hide information?

I. Abstract

The digitisation of courts is often associated with increased procedural transparency and improved access to information. But does digitisation eliminate the information asymmetry or restructure it? As the technology improves, so does the technique of concealing data. Digressing from the traditional form of concealment, digital litigation enables sophisticated forms of non-disclosure. Information is technically disclosed but the manner it is disclosed makes it practically inaccessible or strategically ineffective. This blog aims to identify different tactics, from data flooding to AI-generated submissions, used by litigants for such non-disclosures. It also sheds light on the digitally sophisticated litigants, who are better positioned to exploit these dynamics, further exacerbating inequalities in access to justice. There is a need to realign existing disclosure standards to prioritise intelligibility over formal compliance that ensures that information is not only available but meaningfully usable, accessible and structured within the digital adjudicatory systems.

II. Introduction

Digital courts has been seen as the green initiative of Indian Judiciary to make court paperless and ease accessibility. From e-filing portals to case-management dashboards, digitisation promises to reduce backlogs and make the justice system transparent and efficient. India has digitised over 660 crores pages of court records and has conducted over 3.97 crore hybrid hearings. The focus has been on integration of Artificial Intelligence (AI) based tools into the judicial workflows. Such initiatives have made it easy for the judges to examine pleadings, chargesheets, etc on any device. These innovations produce formal transparency that shows the digital trail of who filed what and when. But formal transparency does not necessarily translate to actual transparency. While the government reports that AI-based modules are being used for record filings, tracking pendency, to reduce human interaction and curbing corruption, there is a rising form of opacity with it. The digitalisation enforces procedural compliances and ensures no document is missing, but can it ensure the relevance and intelligibility of the document produced. This blog aims to explore how digitisation has restructured information flows (i.e., the ways in which legal information is generated, submitted, organised and accessed digitally within the court system) and in the process has created new ways to conceal information. These concealments can occur through various ways like data flooding, metadata manipulation, using AI tools to generate fake submissions, hindering the judiciary process on the interface-level. Litigants can now weaponise data volumes and metadata to hide or dilute important facts. E-filing allows litigants to submit an excessively large volume of documents diluting the relevancy of the information being uploaded and leaving the requesting parties to sift through the data dump.

III. Shift from concealment to design based obfuscation

Data flooding: In a traditional setup, non-disclosure meant refusing to produce a document or delaying the process of its release. Digital documentation introduces a counterintuitive way of doing it by dumping documents. Instead of outrightly evading the document production, litigants choose to overburden the opposite party as well as the judicial system by deliberately uploading large volumes of irrelevant or redundant material into the system with an intention to bury key documents in noise and shifting the cost of sorting to the party who has requested the document. This sheer volume of electronic documents can pose logistical challenges and require careful scoping, early assessment and deduplication. Litigants may flood the record by providing everything to hide something that matters in the absence of structured protocols. In a certain case in Canada, the plaintiff had produced around 4 million documents using an Machine-Learning (ML) tool that is trained to decipher the relevancy of the documents with a precision rate of 81 percent. Interestingly, the court ruled in the favour of the plaintiff after dismissing the defendant’s application to get an amended list of documents. While the court acknowledged that the receiving party should get meaningful, reliable, and complete disclosures for the review purposes, it also noted the lack of an industry standard for precision rates. The standard is not perfection. By dismissing the document dump challenge made by the defendant, the Court risked normalising overproduction as long as the party can invoke proportionality principle and the agreed upon technological protocol. This judgement shows how the bar for compliance is being set very low where a “dilatory and casual attitude to production of documents” must be present. Prioritising procedural pragmatism over catering to the discovery objective of relevant information may result in the requesting party getting stuck with unwieldy and only partially meaningful disclosure.
Metadata manipulation: Digital evidence carries metadata i.e., data about the data like timestamps, author information, file paths, and version histories related to a document. The authenticity of digital evidence, without any supporting metadata, cannot be assessed based on their production alone. Indian courts have started focusing on the when, where and how of the data rather than just the data. Section 63 of Bharatiya Sakshya Adhiniyam, 2023 (BSA) lays down the process for admissibility of such metadata. Further, the Supreme Court has also emphasised the mandatory nature of Section 65B(4) of Indian Evidence Act, 1872 which is the older version of BSA. But metadata has not been given the same leverage as the digital evidence as it is considered a mere technical formality. The documents that are shared with the courts are usually without any version history or emails are produced without any access logs. Sometimes timestamps can be backdated or forward dated or the authors name, or system information might be tampered with.
AI as an evasive tool: AI has been hailed as a tool to reduce information asymmetry by prioritising documents, highlighting relevant precedents, and predicting outcomes. Yet, AI is also being used as a strategic filter as there have been no proper rules or guidance in place, in the Indian judicial system, for disclosure of AI use in litigation. Consequently, this has given rise to the use of AI to generate submissions with fake citations that might look legitimate but no real case exists pertaining to that citation. This is caused due to the hallucinating feature of AI tools where a model generates output that appears legit as they are based on multiple actual incidents but there is no factual existence of such output in itself. Recently, a litigant was fined an amount of Rs 50,000 by Bombay High Court for submitting non-existing judgement. This issue is not just restricted to India but has been a cause of concern for judiciary systems across the globe. A pair of attorneys in New York were met with financial sanctions for relying on ChatGPT which in turn generated bogus case laws. Singapore, in 2024, came out with a guidance on the use of Generative AI tools by court users. The court has a neutral stance on the use of Gen AI tools as it does not prohibit the use of AI but holds the user generating such AI output accountable for such content. The Indian Supreme Court has remarked that such conduct of generating fake submissions amounts to misconduct and cannot be treated as a mere error of law. But there has been no guidelines in place to guide the lawyers how to use AI without transgressing the judicial sanctity.
Interface-level hindrances fudging: Sometimes the design of interface can hinder discovery. If documents are not tagged properly or scanned as non-searchable images, then they become functionally absent despite being formally produced. One of the hurdles that the courts faced in the digitisation efforts was to convert scanned, non-searchable PDFs into machine readable documents. In Metlife Investors USA Insurance Co. v Lindsey, the US court ruled that production of non-searchable PDF files was inconsistent with discovery rules and prevented the opposite party from accessing the metadata and hindered their searching privilege in general. This shows how interface choices can be used to impede the process of justice. While Indian courts have made significant progress in combatting the issue of non-searchable pdfs, the initiatives have varied from state to state and have not been uniformly adopted throughout the nation. For instance, Telangana High Court uses Optical Character Recognition (OCR) software to convert scanned files to editable documents. The Supreme Court has laid down a model rule for e-filing based on which the Delhi High Court issued a circular that mandates the uploading of OCR searchable PDFs under its own e-Filing Rules. Other high courts like Odisha, Maharashtra, Calcutta have also adhered to this practice but there are states like Allahabad who have not adopted the model rules as it already had its own e-filing rule in place. But the rules do not provide any explicit provision for mandatory production of searchable PDFs.

IV. Conclusion

There is no doubt that digitisation has reimagined and transformed our court systems but they have also paved the way for new forms of difficulties and information asymmetries. Physical concealment has been replaced with data flooding, metadata manipulation, etc as discussed above. In order to overcome these difficulties, courts must look beyond infrastructure and adopt regulatory frameworks that address these tactical non-disclosures. There should be guidelines pertaining to uniform AI-use disclosure, protocols for structured and auditable e-discovery like advanced detection of metadata manipulation, a right to intelligible disclosure and proper support mechanism for less resourced litigants in place to ensure that digitisation not only leads to formal transparency but also does not digress from its core objective of access to justice.

Non-Disclosure of data in digital court system : Does tech make it easier to hide information?

Table of Contents

I. Abstract

II. Introduction

III. Shift from concealment to design based obfuscation

IV. Conclusion

SHARE

RECENT ARTICLES

GSTAT’s Real Challenge Is Not Backlog – It Is Transition!

National Judicial Reforms Commission: A proposal we ignore at our peril [Part II]

National Judicial Reforms Commission: A proposal we ignore at our peril [Part I]

We’re not ready for ‘One Case One Data’. Court records are too unorganised for digitisation

Swikruti Mohanty