Deepfakes are synthetic media in which artificial intelligence generates or manipulates video, images, or audio to make someone appear to say or do something that did not occur. They are a digital media dilemma because they enable valuable expression but can also facilitate deception, fraud, and harm.
This topic is part of 3.5 Media and applies to both SL and HL. Deepfakes commonly use machine learning trained on recordings of a person's face or voice. The system identifies patterns and generates media that imitates those features; it does not understand the person or whether the output is truthful.
Deepfakes intensify characteristics of digital media, including rapid distribution, easy reproduction, and convergence between computing, communication, and content.
| Opportunity | Dilemma |
|---|---|
| Film dubbing, visual effects, satire, and creative expression | Viewers may mistake fabricated content for authentic evidence |
| Voice or image reconstruction with informed consent | A person's likeness may be used without consent, weakening control over identity |
| Educational or accessibility applications | Impersonation can support fraud, harassment, or political misinformation |
| Detection tools may improve media literacy | A liar's dividend may allow genuine recordings to be dismissed as fake |
A specific real-world example reported by Hong Kong police in 2024 postdates the 2022 coursebook. System: Fraudsters used deepfake video and audio to imitate company personnel. Example specifics: An Arup finance employee joined a video conference containing imitations of the chief financial officer and colleagues, then transferred about HK$200 million. Impacts and implications: The company suffered financial loss, while businesses face the risk that familiar faces and voices may no longer provide reliable authentication. Concepts: The case reveals unequal power, ethical problems involving consent and deception, and unintended consequences within communication systems.
In an IB response, distinguish established impacts from future implications and evaluate both benefits and risks. Avoid the misconception that all AI-generated media are deepfakes or harmful. Explain a specific example in relation to affected people and communities rather than merely naming it.
