MeitY Advisory Kickstarts a New Regime for AI in India

Restriction on use for prohibited information / content:

“Intermediaries” and AI based “platforms” are obligated to ensure that any AI tool on its “platform” or being used through its computer resource, should not deal with unlawful or wrongful content that threatens the integrity / security of India or its relations with other countries, pornographic material, false or misleading information, infringing material, etc. With the upcoming general elections in India, it is likely that the Government is aiming to curb instances of fake news and misleading information circulated through digital platforms in the country, particularly in the period leading up to the general elections.

Steps to prevent bias in the integrity of electoral process:

“Intermediaries” and AI based “platforms” are required to ensure that their “platform” does not permit any bias / discrimination / threat to the integrity of the electoral process through the use of AI tools. This means appropriate mechanisms may have to be implemented by “intermediaries” to promptly detect and address any such instances, especially when the content is generated through AI tools.

Explicit approval of the Government before using and making available under-testing/ unreliable AI Tools in India:

The Advisory mandates that the “intermediaries” and AI “platforms” take explicit permission of the Government prior to using and making available AI Tools that are un-tested and unreliable. Notably, the foregoing terms are not defined under the IT Act and the Intermediary Guidelines and hence it will have to be seen what interpretation is accorded to them. As previously noted, Government has clarified that the Advisory is aimed at governing “significant platforms” and does not govern startups. However, in absence of a clear-cut definition of “significant platform” under the law, a formal clarification is awaited in this regard. Additionally, the Advisory requires the “intermediaries” and AI “platforms” to affix appropriate labels regarding the possible and inherent fallibility or unreliability of the output generated through such AI Tools, with an aim to clarify to the users that information may not be reliable and correct in all cases. The Advisory suggests that this can be reinforced by the use of “consent pop-ups” to inform the users.

Clear communication to the users about the consequences of dealing with unlawful information:

The Advisory requires the “intermediaries” and digital “platforms” deploying AI to communicate to all of their users, through the terms of services and user agreements, about the consequence of dealing with the unlawful information on its “platform” which may include disabling of access to or removal of non-compliant information, termination or suspension of access to the user account, and punishment under applicable law.

Appropriate labelling of information that may be used as misinformation or deepfake:

The Advisory provides that any intermediary that permits or facilitates synthetic creation / generation / modification of any information (be it text, audio, visual or audio-visual), which can potentially be used as “misinformation” or “deepfake”, should label such information or embed it with a permanent unique metadata or identifier. The purpose of such labelling would be to (a) identify that such information has been modified in the course of dissemination and (b) identify the creator / first originator of such “misinformation” or “deepfake”.

Filing of action taken report:

The Advisory requires all “intermediaries” to ensure compliance with the steps above in the form of an “action taken-cum-status” report within 15 (fifteen) days of the Advisory.