- Deepfake Watch
- Posts
- Deepfake Watch 4
Deepfake Watch 4
A New Deepfake Helpline For Indians | Deepfake Watch
Browser View | February 23, 2023 | Subscribe
A France24 anchor was impersonated using AI to make a fake announcement about an assassination attempt on the French President in Ukraine.
Two friends set out to make better movie dubbing in Polish. Their efforts inadvertently resulted in the infamous Joe Biden robocall (and a plethora of scams).
In Munich, 20 Big Tech companies announced an industry-wide accord to fight deepfakes, while fact-checkers in India made their own alliance to combat the very same thing ahead of elections.
Also, Google is going to use data from Reddit to train its AI models. Yikes?!
Opt-in to receive the newsletter every Friday.
France24’s anchor impersonated
A viral video appeared last week showing a television recording of a news bulletin by France24 on its French-language channel.
The anchor in the video, Julien Fanciulli, announces that French President Emmanuel Macron had postponed his trip to Ukraine after receiving intel on an assassination attempt against him.
It turned out to be a deepfake.
A France24 employee told us that because the video was filmed on a TV screen from a distance, it became increasingly difficult to tell that it is machine-generated.
Former Russian President Dmitry Medvedev also made a reference to this clip in a tweet.
Two dogs podcasting on a mountain
Last week social media was greeted with a flurry of eye-popping visuals.
Pirate ships were battling in a coffee mug, archeologists were excavating plastic chairs, an ant took us on a tour through an ant nest, dogs were podcasting from mountain tops, and a man bowed to a giant cat king in a cathedral filled with cats.
None of these were real - they were made using OpenAI’s latest text-to-video tool called Sora.
Currently available only to OpenAI’s Red Team to evaluate its safety risks, Sora has been a cause for fresh concerns in this big election year. Can OpenAI’s filters be evaded, and can abusive and polarising videos be created by the bad folks of the internet?
We shall find out once the public gets their hands on it.
Movie Dubbing To Scams: The Journey Of ElevenLabs
Two friends from Poland were tired of bad Polish movie dubbing of Hollywood films. So they created ElevenLabs, a voice cloning tool.
It soon got widespread attention from scammers and disinformation factories around the world, and particularly in India.
Decode has detected the use of ElevenLabs in impersonating public figures to promote fraudulent funds, and also create fake audio clips of political candidates ahead of the Madhya Pradesh assembly elections last year.
The same tool was used to create the Joe Biden robocall, where the US President was impersonated to dissuade his supporters from voting in the primaries.
A report by Bloomberg highlighted how the company has been scrambling to catch up to the abuse of their tools by implementing new measures and restrictions.
ElevenLabs had earlier released a ‘speech classifier’ to help detect the use of its tool for voice cloning.
Recently the company stated that it shall disallow the cloning of voices of political candidates heading for presidential and prime ministerial elections. However, according to their press release, this specific restriction would apply to the UK and US, to begin with.
But what happens in India?
A recent report by Al Jazeera highlighted how generative AI is set to play a central role in political campaigning in the country.
The report pointed out how a deepfake video of Bharat Rashtra Samiti leader KT Rama Rao went viral on Congress-run WhatsApp groups, before it appeared on the party’s official X handle.
An AI startup called The Indian Deepfaker has been working with political candidates around the country to help them create highly personalised messages for their electorate, making them address voters by their names.
Divyendra Singh Jadoun, who runs the startup, told Decode last year on how he had self-trained himself on using generative AI to make advertisements, and was eventually approached by political candidates to help with campaign videos.
In a recent conversation with Al Jazeera, he claims to be swamped with “unethical requests” by political clients to create altered videos and audios to target opponents, which included requests of deepfake pornography, which he had denied.
A deepfake helpline emerges
The Misinformation Combat Alliance - a coalition of Indian fact-checking organisations (including BOOM), researchers and media forensic experts have teamed up with Meta to create a WhatsApp helpline ahead of the general elections to combat deepfakes.
Through this helpline, WhatsApp users can send videos, images and audio clips for verification by MCA’s newly launched Deepfake Analysis Unit, or DAU.
The helpline will be accessible to the public from March 2024.
Meanwhile, 20 tech giants met up at the Munich Security Conference and signed an accord to work “collaboratively” to fight the misuse of their AI tools for spreading election-related misinformation.
These efforts include already announced techniques of watermarking images with invisible markers to clarify their provenance, and indicate any alteration.
The Reddit-Google Deal
Just so you know, Reddit just struck a deal with Google for $60 million a year to provide its data to the big tech giant to train its artificial intelligence models.
In return, Reddit will also get access to Google’s Vertex AI, which helps companies improve their search results through machine learning.
Reddit has a bit of a mixed history. It’s known for its diverse and passionate user base, and open nature.
But it is also prone to disinformation and echo-chamber effects, leading to the proliferation of alt-right groups, climate change deniers and conspiracy theories.
How would that affect Google’s AI models remains to be seen.
About Decode and Deepfake Watch
Deepfake Watch is an initiative by Decode, dedicated to keeping you abreast of the latest developments in AI and its potential for misuse. Our goal is to foster an informed community capable of challenging digital deceptions and advocating for a transparent digital environment.
↗️ Was this forwarded to you? Subscribe Now
We invite you to join the conversation, share your experiences, and contribute to the collective effort to maintain the integrity of our digital landscape. Together, we can build a future where technology amplifies truth, not obscures it.
For inquiries, feedback, or contributions, reach out to us at [email protected].
🖤 Liked what you read? Give us a shoutout! 📢
↪️ Become A BOOM Member. Support Us!↪️ Stop.Verify.Share - Use Our Tipline: 7700906588
↪️ Follow Our WhatsApp Channel↪️ Join Our Community of TruthSeekers
Copyright (C) " target="_blank">unsubscribe