Synthetic media: On labelling of AI-generated content.
The requirement of labelling AI-generated imagery is a good start.
Artificial Intelligence (AI)-generated deepfakes, or synthetically generated images and videos, have been proliferating on social media ever since generative AI made photorealistic content as easy as typing a description of a given scene. As this technology developed rapidly over 2024, concerns were raised over electoral integrity and disinformation amplified with the aid of such AI-generated content. While those concerns did not reach the peaks of the feared outcomes, AI “slop” has become pervasive, from low-budget advertisements to high-budget political caricatures. In this context, the Union government’s move to propose mandatory labelling of AI-generated content with an amendment to the IT Rules, 2021, advances the global conversation somewhat in navigating this phenomenon. It is unclear how big the problem of being unable to identify AI slop is for Internet users in India — also the world’s second largest AI user base. Two key factors tilt the scales in favour of acting: as with pre-AI disinformation and misinformation, lies and misleading content may at any time explode into virality, granting them a disproportionate place in the democratic setup; and the technology to create photorealistic content that deceives many people is improving every week. Public personalities have frequently complained, sometimes in court, of the usurpation of their likenesses in the service of unscrupulous projects.
Unlike mandatory smoking warnings at the cinema and on OTT streaming services, an intervention that had little industry buy-in before being imposed on entertainment producers and consumers alike with nothing to reliably show for impact, labelling has been offered up by large social media and AI firms from the very beginning of synthetic content’s increasing accessibility. Meta, Inc. decided last year to label AI-generated content on Facebook, while the Coalition for Content Provenance and Authenticity (C2PA) has brought industry players together for working on “digital provenance” issues, borrowing a concept from the art appraisal world. Subordinate legislation may not have been the ideal instrument to effect this change — the IT Rules already regulate streaming services, social media content removals and the now-prohibited real money gaming industry, all without being tested in Parliament. At some point, these rules need to be explicitly considered by people’s representatives. Technology policymakers have mused throughout the Internet-era that regulation lags behind innovation. The government must be willing to dynamically follow up this proposal with agile action both in terms of relaxing rules that have outlived their utility, and introducing new ones according to requirement.