What is a mitigation strategy for advanced threats in AI?

• A. Continuous monitoring and AI fairness/bias testing
• B. Ignoring bias
• C. One-time safety review
• D. Relying solely on cryptography

Answer: (A)

Continuous monitoring with ongoing AI fairness and bias testing directly addresses how AI systems behave in the real world, where threats evolve and data can drift. By watching runtime performance, data quality, input patterns, and outputs, you can spot anomalies, shifts in behavior, or signs of manipulation as soon as they appear. Pairing that with regular fairness and bias checks helps ensure the system’s decisions remain equitable and less vulnerable to exploitation that could come from biased data or outputs. This combination supports timely updates, retraining, and stronger controls, so risks are managed continuously rather than relying on a single safety review that quickly becomes outdated. Ignoring bias leaves room for harmful outcomes and clever exploitation, while cryptography alone protects data confidentiality but doesn’t guard against unsafe model behavior or biased decisions.