Images of Indian celebrities 'visiting Saif Ali Khan in hospital' are AI-generated

"When Bollywood actors and actresses go to hospital to visit Saif Ali Khan," read a Bengali-language Facebook post published on January 18, 2025.

It shared a collage that appears to show Bollywood superstars Amitabh Bachchan, Shah Rukh Khan and Salman Khan alongside Saif Ali Khan in a hospital bed.

The post emerged after the actor's team announced he was injured during an attempted burglary at his residence. 

Niraj Uttamani, a top official at Mumbai's Lilavati Hospital, told The Hindustan Times newspaper the actor had six injuries, including two that are deeper (archived link).

Image

Screenshot of the false post taken February 13, 2025

The Times of India reported Shah Rukh Khan visited Saif Ali Khan in hospital, although there were no official reports on visits from Bachchan and Salman Khan (archived link).

The collage was also shared in other Facebook posts alongside similar claims.

In fact, the images were made with AI.

A reverse image search on Google led to an Instagram post with four pictures, two of which match the first and second images of the false posts (archived link). 

Shared on January 21, the post's caption reads, "Salman Khan visits hospital to meet Saif Ali Khan. Made with AI."

Another reverse image search for the third photo of the collage led to a Facebook post with four images published on January 18 -- one of the photos matches the third photo from the false post (archived link). 

The caption of the post reads, "Amitabh Bachchan visits hospital to meet Saif Ali Khan. Made with AI".

The Facebook user also posted other photos in the false post between January 17 to 18, with a caption that bears a disclaimer that the images were made with artificial intelligence (archived links here and here). 

Furthermore, the name of an artificial intelligence chatbot developed by xAI "Grok" could be seen at the bottom-right corner of each image. 

Image

Image

Image

Screenshot comparisons of the pictures seen in the false posts (L) and in the original posts, with the 'Grok' watermark highlighted by AFP (R)

Shu Hu, head of Purdue University's Purdue Machine Learning and Media Forensics Lab in the United States, highlighted clues suggesting the images are AI-generated (archived link).

"In the first image, the pupil is absent and in the second image, the patient's gaze angles of the two eyes are inconsistent and his hand is missing, while the fingers of the person on the right appear abnormal," he told AFP on February 19.

Image

Image

Screenshots of the pictures made with AI, with visual consistencies highlighted by AFP

He went on to say there is glare on the glasses in the third image when there is "no indication of light on the front of the face". 

"Additionally, the left eye appears abnormal compared to the right," he added.

Image

Screenshot of one of the pictures made with AI, with visual consistencies highlighted by AFP

Another AI expert Emmanuelle Saliba, chief investigative officer at deepfake detection company GetReal Labs, said the watermarks are enough to prove they were made using Grok (archived link). 

She also said the images have a few signs of being AI-generated, including "the blurred undefined background, the waxy skin and some of the facial features".