YouTube has launched an AI avatar tool for Shorts that allows creators to clone their face and voice and deploy it in videos without filming themselves. The feature is available globally for users aged 18 and older, excluding Europe. YouTube CEO Neal Mohan first teased this in his January 2026 annual letter.

Creators record a short selfie video capturing their face from multiple angles while reading prompted lines aloud. YouTube uses this footage to build a photorealistic avatar powered by Google’s Veo AI models. Each AI-generated clip can run up to eight seconds, and creators can stitch multiple clips together to create longer sequences. Creators can also insert avatars into existing Shorts via the Remix menu. YouTube is pitching this as a productivity tool: creators who cannot film on a given day can let their avatar handle content instead.

YouTube attaches SynthID and Coalition for Content Provenance and Authenticity (C2PA) labels to all avatar content to identify it as AI-generated. Creators can limit or remove remixes of their avatar content. YouTube automatically deletes avatars after three years of inactivity, though deleting an avatar does not remove already published videos.

A YouTube spokesperson said in a statement to MediaNama that since March 2024, the platform has required creators to disclose when realistic content, content a viewer could easily mistake for a real person, place or event, is made with altered or synthetic media. For videos touching on sensitive topics like health, news, elections, or finance, YouTube applies a more prominent label directly to the video player. YouTube has not confirmed whether these measures satisfy India’s SGI labelling requirements under the IT Amendment Rules 2026.

India’s deepfake rules cover this, but only partially: India’s IT Amendment Rules 2026 require intermediaries to deploy “reasonable technical measures to prevent avatar-generated content that falsely depicts a natural person by misrepresenting their identity, voice, conduct, or statement in a manner likely to deceive or constitutes non-consensual intimate imagery,” SFLC.in told MediaNama. YouTube partially satisfies this requirement since only the account owner can create an avatar, and creation requires content that the user shoots and records live on the platform.

However, the rules are harm-based and deception-focused, rather than grounded in consent.

“Even where a person’s likeness or voice is used beyond the scope of their consent, the legal trigger under the Rules may not activate unless the content is also misleading or otherwise unlawful,” SFLC.in said.

As MediaNama reported in March, there is no clear evidence that YouTube has fully implemented the SGI labelling and verification systems required under the IT Amendment Rules 2026, which came into force on February 20. MediaNama has asked platforms, including Google whether they have implemented SGI labelling as mandated under the rules. Google has not responded.

Disclosure labels carry legal weight under Indian law: The IT Amendment Rules 2026 require platforms to prominently label synthetically generated content and embed that label as permanent metadata. Platforms “shall also not enable the modification, suppression, or removal of the label or permanent metadata, including the unique identifier, displayed or embedded.” Platforms that fail to comply lose safe harbour protection under Section 79 of the IT Act, SFLC.in told MediaNama.

If an account is hacked, YouTube may not be liable: Primary liability for misuse of a hacked account falls on the malicious actor under the IT Act. However, if a compromised account produces unlawful content, such as CSAM or non-consensual intimate imagery, YouTube will need to prove that it deployed “reasonable and appropriate technical measures (such as adequate content moderation) to prevent harm” in order to claim safe harbour protection, SFLC.in said.

Non-celebrities have limited recourse on personality rights: India has no codified personality rights law. As MediaNama has reported, the Bombay High Court in March 2026 granted interim relief to Shilpa Shetty in a case that raised whether an AI platform generating a celebrity’s personality on demand without consent can claim safe harbour under Indian law. But legal precedent is currently “limited to celebrities,” SFLC.in said. Non-celebrities would need to approach courts to prove misuse and seek remedies.

A YouTube spokesperson said the platform has expanded its likeness detection tool to all creators in the YouTube Partner Program, allowing them to detect and request removal of unauthorised AI-generated content featuring their face. The platform has also updated its privacy request process to allow anyone, not just creators, to request removal of AI-generated content that mimics their likeness, including face or voice.

The broader contradiction: YouTube’s rationale is that offering this feature in-platform with guardrails is safer than creators turning to unregulated third-party deepfake tools. However, the platform removed 16 AI-generated channels totalling 4.7 billion views in January 2026 for gaming its recommendation algorithm. The only meaningful difference between a faceless AI channel and a creator avatar is who controls it.

Meta already allows creators to build AI extensions of themselves through Meta AI Studio across Instagram, Messenger and WhatsApp. For Indian regulators still tracking platform compliance with the IT Amendment Rules 2026, YouTube’s avatar tool widens the gap between what the rules anticipated and what platforms are now deploying.

Also read:

•  MeitY Scientist Flags ‘Virality’ as Core Risk in Deepfakes Regulation
• How Should the Government have Gone About Deepfake Regulation?
• Explainer: Why detecting deepfakes is a challenging problem

Update, 16th April, 9:58AM: YouTube reached out to MediaNama with a response after this story was published. The responses have been added to the relevant sections of this story.