Deepfakes of the Week

Turkey’s general elections for President took place on Sunday, with the results leading to a runoff taking place in under two weeks. Before voters finished going to the polls, candidates opposing current President Recep Tayyip Erdogan faced two deepfake related scandals.

Kemal Kilicdaroglu, Erdogan’s main challenger in the election, was purportedly shown in a video speaking perfect English on the same day when Kilicdaroglu himself warned supporters of high-tech meddling. Kilicdaroglu will face Erdogan in the runoffs later this month. (Results for the video are below.)

Muharrem İnce, another candidate in the election, had a purported sex tape publicly leak to the internet, which he claims is a deepfake. Critics claim the leak is a stunt to deflect from İnce’s poor performance at the polls. We were not able to scan said video due to lack of direct access and explicit research approval from İnce and his campaign.

What does Reality Defender's platform say?

What does the Reality Defender team say?

This is the highest profile use of deepfakes during an election to date, and it won’t be the last. As deepfakes grow in complexity, scope, and ease of use, more bad actors will use deepfakes to spread political misinformation during the election cycle.

At the same time, more incidents are being blamed on deepfakes that may or may not be deepfakes. (See our newsletter from a few weeks ago on Elon Musk’s faulty deepfake defense.) The only way to tell fact from fiction is not deepfake watermarking, but comprehensive and proactive deepfake detection and scanning on all media, always.

Minnesota Says No to Deepfakes

Minnesota nearly unanimously passed a bill that would make it illegal for individuals to share non-consensual sexual deepfaked content, as well as any content that may influence an election or impact a candidacy. The bill needs to be signed by Gov. Tim Walz and first go through a conference committee, but upon passing, violations could lead to 5 years in prison and up to $10,000 in fines.

Man Arrested for Using ChatGPT

A man was arrested in the Gansu province of China for using ChatGPT to create a news story of a fake train crash. This marks the first known arrest from the usage of AI in China, and comes at a time as the company cracks down on external usage of foreign AI platforms while companies within the country develop their own. 

Google <3s GenAI

Google’s I/O conference this year focused heavily on new additions to their product line and their main search product, including (but not limited to) an experimental implementation of generative searches on Google.com, improvements to Bard, expansion of use cases for their own LLM, implementation of generative content (thanks to a partnership with Adobe) and generative content watermarking. Also of note: a new universal translator that essentially deepfakes the speaker’s lips to work with the corresponding translated audio. You can watch the company’s full keynote here.

More News

• The “leaked” Frank Ocean tracks bought by fans were, in fact, deepfakes. (The Independent)
• Speaking of which, musicians have been using AI at an alarming rate. (Pitchfork)
• The Irish Times has apologized for publishing a generated article featuring false information. (The Irish Times)
• This week featured several notable names and organizations speaking out against the widespread use, abuse, and development of AI, including:

• Stuart Russell, former UL government advisor (CNN)
• Steve Wozniak, Apple co-founder (The Street)
• Various doctors and public health officials in the journal BMJ Global Health (The Guardian)
• Anthropic, a startup composed of ex-OpenAI employees who are making a “safer” alternative to ChatGPT known as “constitutional AI.” (The Verge)

• You will soon be able to order Wendy’s drive-thru via an AI chatbot. (The Wall Street Journal)
• What would AI restrictions in our hometown of NYC look like? (Gothamist)
• A woman created an AI-generated girlfriend based on her likeness for others to use. She is now raking in six figures. (Vice)
• Bestselling book covers now feature generative art. (Twitter)
• Even pro wrestling t-shirts aren’t safe from generative art. (Paste)

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.