Despatch from Washington | International

The rise of ‘deep-fakes’ and threats to truth

Imagine watching a familiar video of yourself on Twitter in which you are, say, speaking at a conference. A few sentences in, you realise that the ‘you’ in the video is saying all sort of crazy things (that you did not say). Soon, all your devices are beeping and chiming and you’ve got hundreds of Twitter notifications as the post goes viral. This is the world of ‘deep-fakes’ — audio and visual-media manipulation taken to a new level altogether, assisted by Artificial Intelligence (AI) and the latest advances in software.

There is no clear definition of ‘deep-fake’, but the term is broadly used to refer to media manipulation rooted in deep neural networks (a form of AI). While the technique can be used to have some harmless fun, it is rife with possibilities of misuse. From creating fake pornographic videos to making politicians appear to say things they did not, the potential for damage to individuals, organisations and societies is vast.

With the 2020 presidential elections looming, lawmakers and other stakeholders in the U.S. are sitting up and taking note. The Pentagon is also involved via its Defense Advanced Research Projects Agency. The Permanent Select Committee on Intelligence held a hearing on challenges of AI and deep-fakes on June 13. “A state-backed actor creates a deep-fake video of a political candidate accepting a bribe with the goal of influencing an election. Or an individual hacker claims to have stolen audio of a private conversation between two world leaders, when in fact no such conversation took place,” said Intelligence Committee Chair Adam Schiff, describing some hypothetical scenarios.

Hardware has become cheaper and more powerful and software more accessible and capable. The incorporation of AI in software has made it “dramatically” easier to manipulate media, Jack Clark, of OpenAI, a research and technology organisation that focusses on the safe use of AI, told the House panel. The ability for social media to make things viral compounds the problem.

A few weeks ago, a video of House of Representatives Speaker Nancy Pelosi was altered to make her appear drunk by manipulating the audio so that her speech sounded slurred. While this was not deep-fake technology at work, it highlighted the danger of such videos.

The Pelosi video also highlighted something else: the different responses across social media platforms, which enjoy considerably freedom from liability. YouTube removed the video and Facebook kept it on there, marked it as “false”, and slowed the speed at which it was distributed.

Relevance for India

The discussion on deep-fakes in the U.S. has much relevance for India. Witnesses providing testimony at the House hearing cited India in their examples to illustrate the dangers of misinformation. For instance, the panel heard how a prominent Indian journalist was at risk after her face was transposed onto a deep-fake pornographic video that went viral.

But individuals are not the only ones at risk, Danielle Citron, a law professor whose expertise includes AI issues, told the House panel. Ms. Citron described a hypothetical scenario of a video that is manipulated so it appears that a company CEO is admitting that the company is insolvent, the night before an IPO. This could derail the company’s IPO.

“There’s no silver bullet. We need a combination of law, markets and really, societal resilience to get through this,” Ms. Citron said.

The Deepfakes Accountability Act, a Bill working its way through the House, seeks to ensure that those creating deep-fake media include with it appropriate disclosures such as watermarks and descriptions; the right of victims to sue creators; and means for victims to protect their reputations when creators cannot be brought to court (e.g., foreign governments). It additionally endeavours to update existing laws around ID theft as well as promote federal research to develop technologies that detect deep-fakes.

Time is of the essence though, and Mr. Schiff summed this up at last week’s hearing. “Now is the time for social media companies to put in place policies to protect users from this kind of misinformation, not in 2021 after viral deep-fakes have polluted the 2020 elections. By then it will be too late.”

(Sriram Lakshman is The Hindu’s Washington correspondent.)

Our code of editorial values

Related Topics
This article is closed for comments.
Please Email the Editor

Printable version | Oct 15, 2021 10:46:59 AM |

Next Story