In a world where technology is advancing at a staggering pace, creating deceptive content has never been easier. Today, with access to artificial intelligence, a few images, and free software, anyone can fabricate events that never happened and make them appear real.
The big question is: Will it be possible to distinguish between reality and artificial creations in the future? Let’s explore the impact of deepfakes, their potential, and the strategies to mitigate their misuse.
What Are Deepfakes and How Do They Work?
The term deepfake comes from the combination of deep learning and fake. It gained popularity in 2017 when a Reddit user began experimenting with this technology. Related terms include face swapping and voice cloning.
By 2018, early examples of deepfakes had already emerged. Director Jordan Peele manipulated a video of former U.S. President Barack Obama, making him "say" things he never actually said. The video went viral, demonstrating just how easy it is to distort reality with this technology.
Other Well-Known Examples:
Pope Francis in a Puffer Jacket.
The "deeptomcruise" TikTok account.
The Princess Diana Scandal: A deepfake of Princess Diana "speaking" about her life and alleged survival circulated on social media, sparking major controversy.
Numerous cases involving high-profile politicians.
The Fake Mark Zuckerberg Video: In 2019, a video surfaced in which Meta CEO Mark Zuckerberg appeared to talk about how Facebook "controls the future thanks to user data." It was a deepfake created to showcase the dangers of this technology.
The Vladimir Putin Deepfake Announcing a Fake Surrender: In 2022, a deepfake video of Russian President Vladimir Putin was released, falsely showing him announcing a surrender in the Ukraine war.
The Fake Volodymyr Zelensky Speech: Also in 2022, a manipulated video of Ukrainian President Volodymyr Zelensky appeared to show him "ordering" his troops to surrender, which was quickly debunked.te desmentido.
The Rise of Deepfakes
The number of fraud incidents involving deepfakes increased tenfold from 2022 to 2023, according to a study by the identity verification platform Sumsub. This growth concerns experts, especially due to its impact on elections and misinformation.
In January 2024, NBC News reported the first known use of an AI-generated deepfake in a presidential campaign: a robocall mimicked President Joe Biden’s voice to influence voters.
How Easy Is It to Create a Deepfake?
It depends on the level of realism desired. For simple videos, apps like iFace are enough. However, sophisticated deepfakes require:
Training an AI model: Analyzing thousands of images, audio clips, and videos of the target person to replicate their features.
Synthesizing the fake content: AI generates images and audio that closely mimic the individual.
As this technology evolves, creating high-quality deepfakes is becoming more accessible, making detection efforts increasingly challenging.
The Fight Against Deepfakes
Regulating such advanced technology is a challenge, but some governments and companies are taking action:
United States: By March 2024, 43 states had introduced or passed laws to regulate deepfakes in elections.
United Kingdom: The Online Safety Bill prohibits the distribution of manipulated images without consent.
European Union: The Artificial Intelligence Act requires deepfakes to be labeled as artificially generated content.
Tech Companies:
Google launched a watermarking tool to identify AI-generated images.
Meta, OpenAI, and 20 other companies signed an agreement to detect and label misleading content.
Adobe, Microsoft, Sony, and Google are working on "content credentials" to authenticate digital material.
OpenAI prohibits the use of public figures in its AI generators.
When truth can be manufactured with the press of a button, the responsibility of distinguishing reality from deception falls on each user. While governments and companies try to catch up, it is crucial to develop critical thinking and question the content we consume.
