AI content norms unimplementable: Social media intermediaries to Meity

Internet intermediaries have told Meity that proposed rules to label synthetically generated information are vague, duplicative, and technologically unviable

The current draft of the proposed amendments for labelling and classification of synthetically generated information (SGI) are “unimplementable” and could “result in large-scale disruptions to various segments of India’s digital economy,” social media and internet intermediaries have told the Ministry of Electronics and Information Technology (Meity).

 

In their submission through the industry body Internet and Mobile Association of India (IAMAI), the companies have said that the existing rules and laws, such as the Information Technology (IT) Act of 2000, the IT Rules of 2021, and the Bharatiya Nyaya Sanhita, already cover all aspects of dealing with such synthetically generated information.

 

“In its current form, the definition of SGI would sweep in routine, good-faith image, audio-visual and text editing that is common across journalism, advertising, education, research, and personal expression,” the IAMAI has said.

 

For example, a separate “cross-cutting” clause for labelling and prohibiting the spread of synthetically generated information could “expand or duplicate obligations, complicate interpretation of due diligence standards, and increase the risk of inconsistent enforcement” across these platforms. Instead of prescribing specific measures for synthetically generated content, the IT ministry should maintain a technology-neutral stance for removing any unlawful content, the IAMAI said.

 

In its submission, the industry body has also said that the current proposed definition for synthetically generated information is “vague and highly subjective”.

 

In the proposed amendments to the IT Rules, the IT ministry has said that any information generated by the use of artificial intelligence and appearing “reasonably authentic” would be classified as synthetically generated and therefore, needs a declaration from the user if such content is uploaded or distributed publicly through any social media or internet intermediary.

 

This, the IAMAI has said, will lead to misclassification, as both intermediaries and creators will face uncertainty.

 

“For example, a colour-corrected photograph or denoised audio track will typically appear authentic to lay viewers, yet it is not misleading in any material sense. Conversely, satire, parody, or clearly stylised content can be ‘inauthentic’ by design without any risk of deception,” the IAMAI has said in its submission.

 

Last month, the IT ministry released a draft of the proposed amendment to the IT Rules of 2021, stating that all social media users who utilise AI to generate and upload content on the internet or any social media platform must declare that AI was used in the process.

 

The IT ministry had given stakeholders until Thursday to comment on the proposed amendments to the IT Rules.

 

In the proposed amendments, the IT ministry had also placed a larger onus on social media and internet intermediaries. It said that all internet intermediaries that allow the use of AI for generating or creating content “ensure that every such information is prominently labelled or embedded with a permanent unique metadata or identifier”.

 

The intermediaries should also ensure that they have the necessary tools and technical measures to verify the accuracy of the declaration made by the user. Further, the intermediaries should also ensure that the user declaration on the use of AI for content generation is “prominently displayed”, the IT ministry has proposed.

 

The IAMAI has, however, said that the inclusion of such a label at this stage is “premature” as the technologies that enable labelling of such content are not yet “reliable, interoperable, or privacy-preserving, and may impose significant burdens without commensurate benefits”. 

AI labelling 

AMAI says existing IT laws already cover synthetic content

 

Draft rules could duplicate obligations, causing enforcement confusion

 

Firms warn labelling tech unreliable, burdensome, and premature 

Proposed rules require AI-generated content to be clearly declared

  Intermediaries asked to verify user declarations and label content