Herald View: Waking up to the menace of deepfake

Deepfake is the new buzzword in town—even media is suddenly all agog. A prominent English weekly put it on its most recent cover: how couldn’t it, being a professed ‘magazine of record’, when even our prime minister has woken up to deepfake’s potential for damage?

The concern is not misplaced, but the flurry of recent statements emanating from government is amusing nevertheless. It tickles because deepfake didn’t arrive yesterday; it only possibly found a new high-profile target, who said ‘whoa!’

On 8 November, days after morphed videos of actors Rashmika Mandanna and Katrina Kaif had surfaced, the Indian government directed social media intermediaries to remove morphed videos from their platforms within 24 hours of a complaint. This had even been mandated in the IT Rules notified in 2021, but if our social media intermediaries were in fact remiss, nobody had pulled them up yet.

No details are available of how many such videos have been identified or how many complaints received, nor is there an explanation why the regulator has taken no action so far.

Within days of the government ‘instruction’, we had the prime minister warning the country of the dangers of deepfake, though he did his best to keep it light, referring as he was to a video showing a lookalike(?) of the prime minister doing the garba.

It wasn’t a deepfake, it turns out, but it did go around a fair bit. No sooner had He spoken than the mighty MeitY (a.k.a. the ministry of electronics and information technology) swung into action—minister Ashwani Vaishnaw, following a meeting with social media intermediaries, assured the media that “companies had agreed on the need for actionable work”.

Not one to be caught sleeping on the job, he added: “We will start drafting regulation today itself and within a short time we will have a new set of regulations for deepfakes”. Hmm. Doctored videos, as most of us know, have been in circulation for many years now.

On more occasions than anyone could possibly count or track, fact-checkers have pointed fingers at the ruling BJP’s own IT cell circulating fake videos, impersonating opposition leaders among other people, and misrepresenting facts, events and contexts.

The minister or his ministry cannot feign ignorance of the red flags that have been raised all too often; if no action was taken, it was presumably because it hadn’t dawned on these mandarins that deepfakes can be a double-edged blade. If they have the potential to embarrass political opponents, they can occasionally target or damage the ruling party as well. This unintended levelling of the battlefield is what all the fuss is really about.

Deepfake videos, with morphed faces and cloned voices, are being reported with a growing regularity every year, as more and more people gain familiarity and acquire the skills of using artificial intelligence (AI) to create deepfakes.

The doctoring of videos requires sophisticated and expensive computing devices, software and loads of data. Sure enough, if there are tools to create deepfakes, there are tools to detect them too. In the era of digital surveillance, the gap between the perpetrator of a crime and the police may have narrowed, but given its potential for inflicting damage or extracting a ransom for holding off, deepfakes have an unimaginably large spread of potential victims.

For example, in the West, there is growing concern over the deep infiltration of deepfakes in the pornography industry, morphing faces of actors in adult films with those of celebrities. The 2020 presidential election in the United States was also allegedly influenced in some part by deepfakes and concerns that we might see worse in the 2024 presidential election have some firm basis.

Back at home, expect our own netas to also ascend a steep learning curve, especially now that the prime minister has indicated there’s nothing funny about garba.

India has over 800 million internet users, a figure estimated to touch 1.2 billion in the next two years. In other words, there is a scary sprawl of potential misuse and victims—not just politicians and rich businessmen, nor only celebrities of all stripes, nor even just the well-heeled with their large or little fortunes to protect but also senior citizens, school and college students, institutions, governments… you get the picture.

So yes, the risk from deepfakes is real, but there is also a risk of overzealous policing and throwing the baby out with the bathwater, the baby being the limitless world of AI possibilities we have glimpsed and must learn to harness.