India has created new rules that state all unlawful content on social media must be removed from platforms within three hours of receiving a court or government order, including AI-generated content.
The new three hours rule is an update to an existing 36 hours rule, meaning social media platforms – like Meta, YouTube, and X – will be under pressure to remove unlawful content much faster.
These amendments are set to take effect on 20th February and will apply for the first time to AI-generated content, such as audio and video, that aims to look real.
This is primarily targeted at deepfakes – a video or image that has been digitally altered to be a person, often used for malicious purposes – but excludes ordinary editing, design or educational content for genuine purposes.
The new rules state that users must clearly label this type of content if it is to remain on social media websites, and must add labels to trace where it came from.
Social platforms will also be required to use tools to detect and stop illegal AI content such as non-consensual content or child sexual abuse.
Critics of the new legislation have warned against the speed at which platforms will need to address unlawful content stating three hours isn’t a feasible amount of time.
A Tougher Stance on AI Content and Deepfakes
India’s move reflects growing global concern about the rapid spread of AI-generated misinformation and manipulated media.
In recent years, deepfake videos targeting politicians, celebrities and private individuals have gone viral in India, particularly during election cycles, where AI deepfakes of politicians were used to manipulate voter perception – raising fears about voter manipulation and reputational harm.
The government has already pressured major platforms to act more decisively against fake media, and this update signals a shift toward stricter enforcement.
By shortening the takedown window to three hours, authorities are placing greater legal and technical responsibility on social media companies to proactively monitor and trace harmful AI content before it spreads widely.
