India's New Deepfake Law: Tech Giants Scramble

India's recently amended Information Technology Rules, effective February 20th, mandate that social media platforms swiftly remove illegal AI-generated content, including deepfakes. Platforms must also ensure that all synthetic media is clearly labeled, potentially by embedding permanent metadata or using other provenance mechanisms. These stringent obligations create a critical challenge for tech companies, as current AI detection and labeling systems are still underdeveloped.

While systems like C2PA aim to provide content credentials, their effectiveness is limited, with metadata often being removed during file uploads. India's new rules prohibit the modification or hiding of such metadata. Companies like Meta, Google, and X have a mere nine days to implement compliant AI labeling systems. Adding further pressure, unlawful materials, including deepfakes, must now be removed within three hours of discovery or reporting, a significant reduction from the previous 36-hour deadline.

Critics warn that these tight timelines could force platforms into widespread automated censorship, potentially leading to the over-removal of content. However, the amendments' specification of provenance mechanisms 'to the extent technically feasible' suggests an awareness of the current technological limitations. This situation presents an opportunity to test and potentially advance the capabilities of AI detection and labeling technologies.