Discuss the ethical issues in using AI to draft official communication. How can the principle of accountability be preserved?

For most citizens, press notes and circulars are their main view of what the government is doing. Public servants already use generative AI to draft them, often without formal approval or clear oversight [1]. This saves effort, but it raises ethical questions about probity, transparency and answerability.

Ethical issues

• Obfuscation at scale: AI cheaply produces smooth, buzzword-heavy "vision language". Citizens have to read more text to find fewer facts, which goes against the ethic of sharing information.
• Diffused responsibility: If an AI-drafted note misleads, it is unclear who answers for it: the officer, the ministry or the vendor. India's AI Governance Guidelines say liability across developers, deployers and users is unclear [2].
• Accuracy: Fluent text can hide wrong figures. Unchecked errors in official claims mislead citizens and the media.
• Erosion of trust: The OECD lists public trust and resistance as risks of generative AI in government [1].
• No fallback for citizens: In the CIC's 2018 transparency audit, only 19% of responding authorities got grade A and 35% got grade E [3]. A vague note often stands in for proper disclosure rather than adding to it.
• Balance: AI itself is not the villain. The G7–OECD Toolkit holds that AI can improve transparency and accountability [4]. The real problem is the incentive to publicise and conceal at the same time.

Preserving accountability

• Named human sign-off: Each release should state whether AI helped draft it and which officer approved it. The UK's Algorithmic Transparency Records are a model [5].
• Public AI-use register: The Netherlands Algorithm Register (2021) and the US federal AI use-case inventory, updated at least yearly, list where government uses AI [5]. Each entry should state the tool's purpose and who answers for its output, so the register does not become box-ticking.
• Binding rules: India's guidelines admit that voluntary frameworks lack legal enforceability [2]. MeitY and DoPT could issue a mandatory procedure for AI use in official communication.
• Use existing RTI machinery: The annual third-party audit of Section 4 disclosures, reported to the CIC [3], could also check whether press releases give measurable targets, actual results and links to source data.
• Checkable claims: Every long-term goal should be paired with this year's target and its result.

AI is only a drafting tool. The ethical burden stays with the person who signs. Disclosure by design and named responsibility can turn AI into an aid to clarity. This serves the right to know under Article 19(1)(a), and the 2nd ARC's view of information as the "master key to good governance" [6].