Evaluate algorithm registers as a tool for public-sector AI accountability. What lessons do they hold for India?

Algorithm registers are public lists of the AI tools a government uses, with each tool's purpose and owner. Examples are the Netherlands' Algorithm Register (2021), the UK's Algorithmic Transparency Records and the US federal AI use-case inventory, which must be updated at least once a year [1]. Registers are a necessary tool for accountability, but they are not enough on their own.

Strengths

• Visibility: citizens, the media and auditors can see where AI shapes public decisions. This is the first step towards challenging those decisions [1].
• Curbs "shadow AI": the OECD finds that public servants often use generative AI without formal approval or clear oversight [2]. A compulsory register forces such use into the open.
• Named responsibility: each entry links a tool to a department that owns it. This supports the G7–OECD view that AI should improve transparency and accountability [3].
• Low cost: a register builds on existing disclosure systems and needs no new regulator.

Limitations

• Box-ticking risk: an entry that gives only a tool's name, with no purpose, data source or responsible officer, reveals almost nothing.
• Disclosure is not redress: a register shows that a tool exists. It gives no remedy to a citizen harmed by a wrong output.
• Self-reporting gaps: hidden, unapproved use escapes the list [2].
• Lag: AI systems change after they are deployed, and a yearly update can fall behind [1].

Verdict: registers make accountability possible but do not deliver it by themselves. Their value depends on compulsory coverage, detailed entries, and links to audit and grievance redress.

Lessons for India

• Make it law: the India AI Governance Guidelines rely mainly on voluntary measures [4]. Public authorities should be legally required to keep a register, for example as a Section 4 RTI proactive-disclosure item.
• Build on audit, not trust: in the CIC transparency audit, only 19% of the authorities that responded got grade A and 35% got grade E [5]. The register should be included in the annual third-party transparency audit.
• Name an officer for each entry: this answers the guidelines' own concern that liability is unclear among developers, deployers and users [4].
• Make it public, not only for regulators: entries should be written in plain language that citizens can read and challenge.

Algorithm registers turn hidden automation into public knowledge, but accountability needs enforceable duties behind them. For India, a compulsory, citizen-facing register built into the RTI and IndiaAI frameworks would make public-sector AI strengthen the citizen's right to information under Article 19(1)(a). Each entry should be detailed, name a responsible officer and be audited regularly.