Deepfakes
What Are Deepfakes?
Deepfakes are audio recordings, photos, videos, and other types of content created using artificial intelligence (AI). Authentic content that has been altered using AI is also referred to as a deepfake. The term has only existed since 2017, partly due to the rapid and continuous development of AI technology.
Deepfakes can be used to publicly humiliate people, spread disinformation, steal someone’s identity, or create pornographic content, also known as deepnudes.
Deepfakes can include:
• Voice cloning or voice manipulation
• Face manipulation, also known as face swapping
How Can I Recognise Deepfakes?
• Poor or inconsistent quality: Are certain parts of the face shown in a higher resolution than the rest of the body?
• Common sense: Does the image or video appear realistic? For example, is the Pope really wearing a puffer jacket?
• Visual errors: AI often struggles to accurately recreate perspective, realistic lighting conditions, shadows, and colour gradients.
• People: Facial expressions, gestures, speech, transitions between the head and neck, and limbs may be displayed incorrectly by generative AI. Examples include six fingers instead of five, an unnatural or empty stare, or distorted body parts. People in AI-generated videos may also blink very rarely or not at all.
• Reverse image search: A reverse image search can show whether an image has appeared elsewhere and in which contexts it has been used. It may even help you find the original image.
• Speech: AI-generated audio is often free of mistakes, accents, and dialects. In an interview involving several participants, for example, it may be unlikely that everyone speaks flawless standard German. Unnatural emphasis, rhythm, or intonation may also indicate that an audio recording has been manipulated.
• When in doubt: Use AI detection tools such as DeepFake-o-meter, Deepware.ai, or Sensity.ai.
Why Can Deepfakes Be Dangerous?
Today, controversial or shocking media can spread within a very short period of time. Users may discover what appears to be breaking news and share it without verifying whether it is genuine. Deepfakes can exploit this behaviour to deliberately spread false information.
Large-scale disinformation campaigns can create resentment, division, and tension within society. They may divide communities and, in serious cases, pose a threat to democracy.
Example: In March 2022, shortly after the beginning of Russia’s full-scale invasion of Ukraine, a video appeared that supposedly showed Volodymyr Zelenskyy calling on Ukrainian soldiers to lay down their weapons and surrender.
Fraud and social engineering involving deepfakes are no longer merely theoretical scenarios. Well-known scams, such as impersonating a family member in an emergency, can be made even more convincing through deepfake technology.
A few clear photos or a clean voice recording, which may now need to be no longer than 30 seconds, can be enough to imitate a person’s face or voice. Criminals could then call that person’s bank and attempt to authorise a large transfer to their own account. They could also impersonate a relative and ask a grandparent for money.
Example: In February 2024, an employee in Hong Kong transferred approximately 25 million US dollars after being instructed to do so during a video call that appeared to include the company’s chief financial officer. It was later discovered that every other person in the call was a deepfake.
https://edition.cnn.com/2024/02/04/asia/deepfake-cfo-scam-hong-kong-intl-hnk
These are not the only risks and dangers associated with deepfakes. It is also important to recognise how easily deepfakes can now be used for bullying, harassment, defamation, and public humiliation.
For this reason, people should take steps to make it more difficult for others to create deepfakes using their face or voice.
What Can Be Done About Harmful or Fraudulent Deepfakes?
• Report the deepfake and the account responsible for distributing it to the operator of the social network or platform.
• The internet is not a lawless space. Deepfakes may be relevant under criminal law, for example under the German criminal offences relating to insult and defamation pursuant to Sections 185 and following of the German Criminal Code. Harmful deepfakes can and should therefore be reported to the police.
• Do not share, like, comment on, or otherwise interact with the deepfake.
• Limit the visibility of your photos, videos, and audio recordings on social media. This can make it more difficult for others to create deepfakes using your face or voice.
