"Generative AI can either deepen or dilute transparency in government communication." Critically examine with reference to India.

Transparency in government communication means citizens receive accurate, verifiable information on State action. Generative AI (GenAI) is an amplifier: the G7–OECD Toolkit holds that AI can improve transparency and accountability [1], yet unchecked use in drafting press notes can multiply vague, buzzword-heavy "vision language".

How GenAI can deepen transparency

• Plain language at scale: dense circulars, notifications and budget papers can be simplified and translated into Indian languages.
• Filling disclosure gaps: a CIC transparency audit (2018) found over 85% of authorities did not disclose budget, programme or "publicity and public interface" details under Section 4, RTI Act [2]; AI can automate such proactive disclosure.
• Responsiveness: AI assistants can answer citizen queries, advancing the 2nd ARC's view of RTI as the "master key to good governance" [3].

How GenAI can dilute transparency

• Obfuscation made cheap: models churn out fluent, generic text; notes heavy on distant 2047 goals and light on current numbers become harder to scrutinise.
• Shadow use, blurred accountability: the OECD finds public servants using GenAI often without formal approval or oversight [4], so no one knows who verified a claim.
• Weak legal backstop: MeitY's India AI Governance Guidelines (2025) are voluntary and lack legal enforceability. They also route transparency reports confidentially to regulators, not citizens [5].
• No public register: unlike the Netherlands' Algorithm Register or UK Algorithmic Transparency Records [6], India has no public list of government AI uses.

Assessment and way forward The root cause is the incentive to publicise achievements and hide shortcomings, not the tool. AI removes the effort that once limited vagueness, and it hides who wrote the text. Hence, regulate rather than ban:

• Disclosure label on every PIB/ministry release, stating whether AI was used and naming the approving officer.
• Extend DoPT's annual third-party transparency audit [2] to check whether releases give measurable targets and source data.
• A public government AI-use register under the IndiaAI framework, with real content rather than box-ticking [6].

GenAI therefore cuts both ways, and its effect depends on the incentives around it. Disclosure, human sign-off and data-linked claims can make it an ally of the citizen's right to know under Article 19(1)(a) and of SDG 16.10 on public access to information.