Artificial intelligence is now deeply embedded in everyday decision-making. From loan approvals and recruitment screening to medical diagnostics and public services, AI systems increasingly influence outcomes that directly affect individuals. As reliance on these systems grows, so does a fundamental concern. The question is, how can decisions made by machines be understood, questioned, and held accountable. This concern has led to the rise of explainable artificial intelligence, commonly known as XAI. At the same time, the companies developing these systems rely heavily on trade secret law to protect their technology. The interaction between these two forces has created a new legal and policy challenge.
Trade secrets have become the most practical form of Intellectual Property protection for artificial intelligence. Copyright law offers limited coverage, typically extending only to source code and not to the underlying logic or learned behaviour of AI systems. Patent protection, while available in theory, often requires public disclosure and struggles to accommodate rapidly evolving or data-driven models. In contrast, trade secret law allows companies to protect valuable information such as model architecture, training methods, datasets, and optimisation techniques, provided they remain confidential and commercially valuable. For many AI developers, secrecy is not just a strategic choice but a necessity for maintaining competitive advantage.
Explainable AI emerged as a response to the growing opacity of machine learning systems. Many modern AI models, particularly those based on deep learning, operate in ways that are difficult even for their creators to fully interpret. XAI seeks to address this by making AI decisions more understandable to humans. This does not always require revealing technical details. Often, it involves explaining which factors influenced a decision and how those factors were weighed. Explainability has become increasingly important in legal, ethical, and regulatory discussions, especially where AI systems affect rights, opportunities, or access to essential services.
The tension between trade secrets and explainability becomes clear when AI systems are used in critical decision making. Individuals affected by an automated decision may seek explanations that go beyond general statements and require meaningful insight into how conclusions were reached. Companies, in response, often resist disclosure by invoking trade secret protection. While some explanations can be provided without revealing proprietary information, there are situations where simplified or surface-level explanations are insufficient. In such cases, secrecy can limit the ability to assess whether an AI system is fair, accurate, or lawful.
To address this challenge, many AI developers rely on explanation tools that generate reasons for decisions without exposing the underlying model. These explanations can be useful, but they are not always reliable. They may simplify complex processes or fail to reveal deeper structural issues such as bias or systemic errors. As a result, concerns remain about whether these approaches genuinely promote transparency or merely offer an appearance of accountability while preserving secrecy.
Trade secret protection has never been absolute and has always been subject to limitations where public interest demands disclosure. Legal systems are now struggling to balance innovation incentives with the need for transparency. In the European Union, this balance is set out in the EU Artificial Intelligence Act, which requires high-risk AI systems to be transparent and capable of explanation. India, on the other hand, does not yet have a dedicated AI law or a specific statute governing trade secret. In India, questions around secrecy and transparency in AI are therefore likely to be addressed through existing legal frameworks, including common law principles protecting confidential information, constitutional requirements of fairness and non-arbitrariness, and transparency obligations under the Digital Personal Data Protection Act, 2023. Although Indian courts have not yet directly considered disputes involving explainable AI and trade secrets, the growing use of AI in areas that affect individual rights suggests that courts will increasingly be required to strike this balance through judicial interpretation rather than through explicit legislation.
The courts and regulators are beginning to explore ways to reconcile confidentiality with accountability. This includes allowing limited disclosure to regulators or courts under protective conditions, while ensuring that affected individuals receive clear and understandable explanations. There is also growing emphasis on designing AI systems that are interpretable from the outset, reducing reliance on secrecy when explanations are required.
The rise of explainable AI requires us to rethink how secrecy works in a world where algorithms increasingly affect people’s lives. As AI systems move beyond improving efficiency and begin making important decisions, secrecy can no longer be treated as the normal or unquestionable. Ultimately, the question is not whether AI systems can remain trade secrets, but under what conditions secrecy is justified. In high-risk contexts, explainability is not merely a technical preference, it is a legal and ethical necessity. Reconciling these competing imperatives will be one of the defining Intellectual Property
By Ananya Reghu
