Finding the Truth in Deepfake Advertising

Deepfake of the Week

Martin Lewis, noted financial journalist and broadcaster, took to Instagram last week to hawk an app and make wild claims about Elon Musk’s involvement. Since then, we’ve learned that Martin Lewis himself had nothing to do with this ad, nor did he ever record the semi-passable video of him speaking.

This instance is the most recent and highest profile illicit use of deepfakes in online advertising, which deservedly drew the ire of Lewis himself.

What does Reality Defender’s platform say?

What does Reality Defender’s team say?

Reality Defender Co-Founder and CEO Ben Colman spoke with Channel 4 News in the UK on the dangers of malicious deepfakes and how ads like this one can fool even the more keen-eyed viewers.

Louisiana Criminalizes CSAM Deepfakes

Louisiana Senate Bill 175 has criminalized "unlawful" deepfakes — synthetically created materials "Depicting a minor engaging in sexual conduct." Creating or possessing this material can lead to a prison sentence of up to twenty years. This is the latest state-led action against deepfakes, and the most recent bill against CSAM.

Reality Defender in Rest of World

Reality Defender Co-Founder and CTO Ali Shahriyari was recently featured in Rest of World as part of our ongoing industry-wide work with Witness’ Deepfake Rapid Response Force. Read more about our analysis of a political deepfake audio clip that widely circulated the internet.

Artists Sue OpenAI and Meta

Comedian Sarah Silverman and authors Christopher Golden and Richard Kadrey have sued OpenAI and Meta, claiming the companies trained their LLMs on copyrighted material hosted on illicit book download sites. Silverman was able to reproduce parts of her book, Bedwetter, through queries on ChatGPT, while the trios’ work were allegedly available wholesale in Meta’s LLaMA training datasets.

More News

• Labor union IATSE laid out their principles for AI use in the entertainment industry, emphasizing research, education, advocacy, organizing, and collective bargaining while planning to identify and train high-risk members. This comes at a time when the Writer’s Guild is on strike and the Screen Actor’s Guild may soon follow. (Variety)
• Yuval Noah Harari, author of Sapiens, believes those creating AI models and content pretending people should face tough criminal sentences, similar to those who counterfeit currencies. (The Guardian)
• Google’s privacy policy now states it will use public data to train AI models. (Engadget)
• One hundred and fifty executives are hitting back at the EU's AI regulations, claiming the rules can negatively impact competition. (The Verge)
• Shutterstock, like Adobe before it, will offer legal protection for its Enterprise customers using the platform’s AI-generated images. (VentureBeat)
• Reddit has banned AI-generated pornography with the exception of content featuring “fictional people or characters.” This move comes after the company’s recent API changes, which forbid existing third-party apps from accessing adult materials (which are, in fact, a large part of Reddit’s content). (Reddit)
• The gaming industry is starting to implement AI during production. (FT)
• Baidu announced their improved LLM, Ernie 3.5. (Baidu)
• UK research universities finally have a collective plan on how to deal with AI. (Quartz)
• A profane, never-ending, and AI-generated Trump vs. Biden debate is live streaming on Twitch (Kotaku)
• ChatGPT actually lost active users. (Ars Technica)
• How is AI harming the planet? (The Markup)

Thank you for reading the Reality Defender Newsletter. If you have any questions about Reality Defender, or if you would like to see anything in future issues, please reach out to us here.