Polity & GovernancePriority2025-10-10

ECI Mandates Labelling of AI-Generated Campaign Content

Why focus: Tests ECI's Article 324 powers to extend Model Code of Conduct to deepfakes — high-yield GS2 Polity application question.

In News

What Happened

On October 24, 2025, the Election Commission of India (ECI) issued a strict advisory mandating political parties to prominently label all AI-generated or synthetic content used in their digital campaigns. The directive requires disclaimers to cover at least 10% of visual media and mandates the removal of deceptive deepfakes within three hours of detection. This was enforced using the ECI's powers under Article 324 of the Constitution to maintain a level playing field.

Why It Matters

The rapid spread of hyper-realistic deepfakes threatens democratic integrity by misleading voters and distorting political narratives. By mandating transparency rather than outright censorship, this policy establishes a global standard for integrating artificial intelligence ethically into electoral processes.

Background

History & Context

The regulation of AI in elections has evolved rapidly due to the surge of deepfakes during the 2024 Lok Sabha polls. The ECI first addressed this on May 6, 2024, asking parties to refrain from using distorted content. Concurrently, the Ministry of Electronics and Information Technology (MeitY) issued multiple advisories under the IT Rules, 2021, emphasizing the need for metadata tagging of synthetic media. On January 16, 2025, the ECI issued a preliminary advisory for labelling, which lacked specific parameters. The October 2025 mandate consolidated these efforts, aligning with draft rules from MeitY to create enforceable, quantitative standards for transparency.

What Changed

▶
Mandatory Visual Labelling: Labels like 'AI-Generated' or 'Synthetic Content' must now cover at least 10% of the screen area for visual media.
▶
Audio Labelling: For audio clips, the disclaimer must be announced during the first 10% of the clip's duration.
▶
Placement Standards: Watermarks and labels on video content must be prominently placed at the top of the screen.
▶
Metadata Disclosure: Every piece of AI-generated content must explicitly state the creator or responsible entity's name in its metadata or caption.
▶
Rapid Takedown Mechanism: Political parties must remove deceptive or non-compliant AI content from official handles within three hours of notice or reporting.
▶
Record Maintenance: Parties are now required to maintain detailed internal records of all AI-generated materials, including timestamps and creator identities, for ECI verification.
▶
Citizen Reporting Integration: Misuse of AI content can now be officially reported by citizens through the C-Vigil app, which aims for a 100-minute resolution timeframe.

What Did NOT Change

The ECI did not issue a blanket ban on the use of AI in political campaigns, allowing parties to continue using technology for data analytics, translations, and ethical voter outreach. Furthermore, existing statutory frameworks like the 48-hour silence period under Section 126 of the Representation of the People Act, 1951, remain unchanged but are now explicitly interpreted to cover synthetic media.

Prelims Angle

NCERT Connection

This connects directly to Class 11 Political Science (Indian Constitution at Work), Chapter 3: 'Election and Representation'. The chapter discusses the role of the Election Commission of India under Article 324 in ensuring free and fair elections. The mandate on AI labelling is a modern application of the Model Code of Conduct, showcasing how the ECI adapts its constitutional powers to counter new technological threats that could disrupt the 'level playing field' among candidates.

Common Misconceptions

✗ The ECI has completely banned the use of AI and deepfakes in political campaigns.

✓ The ECI permits the use of AI and synthetic media provided it is prominently labelled, transparent about its source, and does not spread misinformation.

Headlines often use terms like 'ECI cracks down on deepfakes', leading people to conflate regulation and labelling with an outright ban.

✗ The new rules require the ECI to pre-approve all AI-generated content before publication.

✓ There is no pre-censorship or pre-approval mechanism; the rules rely on post-publication labelling, metadata disclosure, and a 3-hour takedown window for violations.

Media Certification and Monitoring Committees (MCMC) require pre-certification for television/radio ads, making people assume digital AI content falls under the exact same pre-approval net.

Practice Questions

Q1

How Many Correct

Consider the following statements regarding the Election Commission of India's October 2025 advisory on AI-generated content: 1. Labels indicating 'Synthetic Content' must cover at least 10% of the visible area for video materials. 2. Political parties are mandated to remove non-compliant deepfakes from their official handles within 24 hours of receiving a notice. 3. The ECI derives the authority to enforce these guidelines under Section 126 of the Information Technology Act, 2000. How many of the statements given above are correct?

Q2

Match the Following

Match the following regulations/mechanisms with their corresponding features in the context of electoral technology rules: List I (Mechanism/Provision) A. Article 324 of the Constitution B. Section 126 of RPA, 1951 C. C-Vigil App D. 3-Hour Rule List II (Feature) 1. Application of the 48-hour silence period to synthetic media 2. Mandatory timeline for parties to take down deceptive AI content 3. Constitutional backing for ECI to enforce the Model Code of Conduct 4. Platform for citizens to report unlabelled AI campaign material Select the correct code:

Q3

Assertion & Reason

Assertion (A): The ECI's 2025 advisory strictly prohibits political parties from using artificial intelligence to translate campaign speeches into regional languages. Reason (R): The ECI considers the use of technology to alter or modify original information as a deep threat that inherently contaminates the level playing field. Select the correct answer from the codes below:

Up next