Which items are examples of emerging AI risks?

• A. Automatic code generation.
• B. Deepfakes and data poisoning.
• C. Cloud service billing.
• D. Keyboard input latency.

Answer: (B)

Emerging AI risks are threats that arise from new or evolving capabilities of AI, especially how these abilities can be misused or exploited to cause harm. Deepfakes illustrate this clearly: as generative AI tools become more accessible and realistic, it becomes easier to produce convincing misleading media, which can be used for impersonation, deception, or manipulation on a broad scale. Data poisoning is another key risk, where attackers subtly contaminate training data or feedback loops to degrade a model’s performance, cause misclassifications, or embed backdoors. Together, these risks reflect how advancing AI can create new avenues for harm to trust, safety, and integrity in AI systems.

Automatic code generation does raise safety and security concerns, but the phenomenon being tested—emerging AI risks in the sense of new, broad threat categories—is more directly illustrated by deepfakes and data poisoning. Cloud service billing and keyboard input latency are not AI risk types; billing is a cost/ops issue and latency is a performance issue.