In the context of AI bias, many AI models are ______. I. Deepfakes II. Black…

Concept

AI systems can be classified along separate axes. Interpretability concerns whether a model’s internal decision logic can be traced and explained; an opaque model is described as a “black box.” Synthetic-media generation concerns outputs such as deepfakes, while k-anonymity is a privacy technique for released datasets. These categories are not interchangeable.

Application

The blank asks what many AI models are in discussions of AI bias. Statement II, “Black boxes,” describes the model itself: its internal decision-making can be difficult to inspect, which also makes biased patterns harder to diagnose. Statement I names synthetic media produced by AI, and statement III names a dataset-anonymisation technique; neither describes the model itself.

Contrast

• “Only I” selects deepfakes alone, which are generated media rather than a property of a model’s decision process.

• “I, II and III” combines model opacity, synthetic-media output, and data privacy even though they belong to three different categories.

• “Only I and III” combines the two concepts that do not describe a model’s internal decision process and omits black-box opacity.

Result

Therefore, only statement II fits the blank, so the answer is “Only II.”