Indian Journalists Battle Election Deepfakes Amid Rising AI Misinformation
As India gears up for major elections, journalists across the country are facing a new and growing threat: AI-generated deepfakes that blur the line between fact and fiction. In late 2023, Karen Rebelo, deputy editor of the Mumbai-based fact-checking outlet Boom Live, uncovered a viral deepfake that underscored the escalating challenge.
The clip in question showed Kamal Nath, former chief minister of Madhya Pradesh, allegedly pledging to end the state’s popular Ladli Behna welfare program if his party won the upcoming elections. The program, a flagship initiative of the ruling Bharatiya Janata Party (BJP), is widely favored by voters—making Nath’s supposed promise seem not only out of character but politically damaging.
“There could be only two possibilities,” Rebelo said. “Either you have the world’s best impersonator, or you’re using a voice clone.” Convinced it was an audio deepfake designed to influence the election, Rebelo and her team at Boom Live set out to verify the clip. But traditional fact-checking tools and strategies weren’t enough.
“You can no longer solely rely on your human skills as a fact checker,” Rebelo explained. The rise of generative AI has enabled the creation of increasingly realistic fake videos and audio clips, complicating efforts to detect disinformation. Fact-checkers are now turning to forensic analysis and advanced AI detection tools to uncover signs of manipulation.
But many newsrooms in India, including Boom Live, struggle with access to reliable detection tools. Rebelo described a frustrating search for trustworthy AI audio analysis platforms, much of it involving free online tools that produced inconclusive results.
Ultimately, it was a cold email to Siwei Lyu, a deepfake expert and professor at the University at Buffalo, that broke the case open. Lyu agreed to conduct a forensic analysis of the audio. His findings confirmed what Rebelo had suspected: the clip of Kamal Nath was a fake. Further analysis uncovered three additional deepfakes targeting politicians involved in Madhya Pradesh’s high-stakes state elections.
Boom Live published its investigation exposing the manipulated audio clips on January 18, 2024. By then, the state elections had concluded more than a month earlier, but the revelations underscored a growing threat to electoral integrity.
As India prepares for its next round of elections, journalists like Rebelo are bracing for a wave of synthetic media designed to mislead voters. Their efforts highlight both the critical role of independent journalism and the urgent need for better detection tools in the fight against AI-driven misinformation.
