Which data-related risk is critical to address to prevent AI security issues?

• A. Training data integrity.
• B. Keyboard layout.
• C. File name conventions.
• D. Electric bill cost.

Answer: (A)

Training data integrity is essential because a model’s behavior comes directly from what it learns during training. When the data is accurate, complete, and unaltered, the model can generalize properly and make secure, trustworthy decisions. If data is corrupted, biased, or poisoned, the model can learn incorrect patterns, produce biased or unsafe outputs, or even leak information, creating security vulnerabilities. Protecting data integrity means ensuring provenance, validating data quality, using secure data pipelines, and verifying data with checksums or versioning so the model isn’t trained on manipulated or low-quality information. The other options don’t influence how the model learns or guards against data-related attacks, so they don’t address the core security risk in AI training.