BJP used ‘deepfake’ videos in Delhi campaign: Report
NEW DELHI: The Delhi Bharatiya Janata Party (BJP) used videos produced with “deepfake” technologies during its recent election campaign in the national capital, according to a report that has now triggered concerns over the potential for election malpractice in the future.
To be sure, the report does not accuse the BJP of malpractice. Nor are the videos themselves a problem. But the report highlights the power of this technology.
Deepfake refers to media – audio or video – created using the artificial intelligence technique of deep learning. Such media can be entirely fictional and can range from portraying a person carrying out an activity they never did or speak words that they never spoke.
According to a report by Vice, the BJP’S Delhi unit use a communications company to produce at least two videos of the party’s leader Manoj Tiwari to show him speaking in Haryanvi and English. They were an alteration of a video made in Hindi, in which Tiwari spoke about a completely different issue from what is seen in the other two videos. Apart from the audio being overlaid in an entirely different language, the lip-synching was tweaked to make it seem as if Tiwari was indeed speaking.
A Delhi BJP representative said the party had not specifically hired any firm and that the videos were made after it was shown a sample. “Many agencies come and display or talk about their products. One of our team members had come across a video of Tiwariji in Haryanavi language so we circulated that in our internal groups. It was felt that if the video had been in English, it would have been better so one of our team members asked for the English video which those guys gave us,” said Delhi BJP’S co-incharge for media and IT Cell, Neelkant Bakshi.
“We had not tied up with any of such agencies and if used positively the technology definitely sounds good,” he said.
According to researchers specialised in tracking deepfakes, the video was not the first time such content has been circulated in India but it was unprecedented in how professionally it was produced. “It is certainly novel because deepfakes have not been seen as such in election campaigns in India,” said Henry Ajder, head of threat intelligence at Deeptrace Labs, an Amsterdam-based cyber security firm.
The BJP videos, while inoffensive, introduce a new tool for political parties. “Once you introduce such technology in the common political tool box, you are changing the way campaign tools can be used in the future,” Ajder said, adding that it opens the potential for more malicious use.
Ajder said a team at Deeptrace had seen the Tiwari video and found signs of it being a deepfake.
“We have seen examples in the Indian context in the past where deepfakes have been used to target politicians by showing them in pornographic videos and to tarnish reputation of journalists,” he said.
Indian legal experts said the video highlights the gaps in Election Commission’s current mechanisms to deal with social media content if it falls afoul of laws.
“The issues in this case are similar to the ones we see in the case of other misinformation: they are hard to trace, third parties can get away with funding it – thereby bypassing spending limits; and the regulatory mechanism rests on a voluntary code of ethics agreed upon by internet and telecommunications industry bodies,” said Apar Gupta, lawyer and executive director of the Internet Freedom Foundation.
“The EC does not seem to be responsive to threats from such technologies. Its strategy involves relying on the Internet and Mobile Association of India (IAMAI) instead of taking action against companies, third parties or the candidates,” he added.
Representatives of EC did not respond to requests for a comment.
According to the Vice report, BJP collaborated with Chandigarh-based advertising firm The Ideaz Factory. The company did not respond to HT’S request for a comment.
According to Deeptrace, there were nearly 15,000 deepfake videos online last year with 96% of them being pornographic.
Ajder said the there has been a consistent rise in such content, “which shows the growing commodification of the technology”. For now, detection of such fake content is nearly impossible without technical help. “There has to be a platform level approach. Detection is possible but it’s becoming increasingly challenging,” Ajder added.
(With inputs from Ashish Mishra)