ECP proposes rule amendments to curb AI misuse during polls

Pakistan’s Election Commission has proposed amendments to its 2023 codes of conduct for political parties, media organisations, and election observers. The measures aim to regulate AI-generated election content, stop deceptive deepfakes, and improve transparency in political communication. The ECP has invited public and stakeholder feedback by Nov 20.

For media organisations, the draft would require clear disclosure whenever AI-generated or AI-modified content is broadcast, published, or released. Media outlets could still use AI to prepare, translate, present, or analyse election material. However, they would remain responsible for the material’s authenticity and accuracy.

The rules would prohibit false, defamatory, or ridiculing deepfakes involving political leaders, candidates, election officials, judges, armed forces personnel, and other state officials. Regulators would monitor compliance, while media houses could be ordered to remove misleading content or label it correctly. The proposals are not yet final.

A deepfake is synthetic media created or altered with AI so that a person appears to speak, act, or appear differently from reality. It can combine a person’s likeness with invented words or actions. The result may look or sound authentic even though the person never made the image, voice recording, or statement.

For example, software can place a candidate’s face onto another video, generate a voice resembling the candidate’s, or produce a video showing the candidate apparently making a false claim. The ECP draft specifically refers to fabricated images, voices, and statements involving political leaders and other officials. It bans such content when it is false, defamatory, or intended to ridicule.

The article focuses on election safeguards rather than technical details of how deepfakes are made. Its proposed response is disclosure for legitimate AI use and prohibition of harmful fabricated content. Media organisations would also remain responsible for accuracy and could be ordered to remove or relabel misleading material.

The ECP’s proposal contains 12 amendments aimed at curbing AI misuse during elections. According to sources cited in the article, the changes would affect three election-related codes of conduct. They cover media organisations, political parties and candidates, and election observers.

The reported breakdown assigns five changes to the media code, five to the code for political parties and candidates, and three to the observer code. Those figures total 13, while the article describes the overall proposal as 12 amendments. The article does not explain the apparent difference, so both figures are reported as presented.

The proposed changes are intended to regulate AI-generated content, prohibit deceptive deepfakes, and increase transparency in election communication. The ECP sought public and stakeholder feedback by Nov 20. This means the proposals were still undergoing consultation rather than operating as final rules at the time described.

The proposed rules would require media outlets to tell audiences when election-related content was generated or modified through AI. Disclosure would be needed when such material is broadcast, published, or released. The goal is to make the technology’s role visible instead of allowing audiences to mistake altered material for wholly human-produced reporting.

The draft still permits media organisations to use AI for preparing, translating, presenting, or analysing election-related content. For example, an outlet could use AI to translate a political report, but it would need to disclose that use. The organisation would remain responsible for the authenticity and accuracy of the final material.

Regulators and the ECP could also direct media houses to remove misleading or undisclosed AI content, or relabel it correctly. The article says undisclosed or unlabeled AI-generated media would be monitored. The proposed disclosure rule therefore combines transparency with continuing editorial responsibility and possible corrective action.

Three authorities would monitor undisclosed or unlabeled AI-generated media content: Pemra, the Social Media Protection and Regulatory Authority, and the National Cyber Crime Investigation Agency. They would report violations to the ECP. The proposal therefore assigns monitoring responsibilities beyond the Election Commission itself.

If a media house carried misleading or undisclosed AI-generated content, a regulatory body or the ECP could direct it to remove the material promptly. Alternatively, the outlet could be required to label it correctly. The draft also says appropriate action would be taken when organisations fail to comply with the proposed requirements.

The article does not specify the exact penalties or enforcement timetable. It describes the amendments as proposals and says the ECP requested public and stakeholder feedback by Nov 20. Their final wording, enforcement process, and consequences would therefore depend on what happens after consultation and adoption.

Deceptive AI-generated political content can make false images, voices, or statements appear genuine. As a general democratic principle, that can confuse voters about what candidates said or did. It may also shift attention from real election issues and make reliable information harder to identify. These broader consequences are not detailed in the article, but they explain why authenticity matters.

For example, a fabricated voice message could falsely appear to come from a candidate. If voters encounter it without a label or correction, they may judge the candidate using information that is not authentic. The ECP proposal addresses this risk by banning certain false, defamatory, or ridiculing deepfakes and requiring disclosure of AI-generated or modified content.

The proposed monitoring system would involve Pemra, the Social Media Protection and Regulatory Authority, and the National Cyber Crime Investigation Agency. They could report violations to the ECP. Media houses could then be ordered to remove misleading material or label it correctly, supporting fairer access to election information.

Democratic elections depend on voters receiving authentic information because voting choices are shaped by what people believe candidates said, did, or propose. If content is altered or falsely presented, voters may be unable to make informed comparisons. This principle is broader than the article, but it underlies the ECP’s focus on authenticity, accuracy, and transparency.

A clearly labeled AI translation, analysis, or presentation lets audiences understand how technology was used. A fabricated video or voice message presented as genuine does the opposite. The proposed rules respond by requiring disclosure of AI-generated or modified media and prohibiting false, defamatory, or ridiculing deepfakes involving candidates and officials.

Trust also requires accountability after publication. The proposal would keep media organisations responsible for the accuracy of AI-assisted material. Regulators and the ECP could require misleading or undisclosed content to be removed or relabeled. Together, these measures are intended to support transparent election communication and informed participation.