CoreQ Academy2.5 Given a scenario, implement monitoring and auditing for AI systems.
Monitoring AI means watching what goes in, what comes out, how much it costs and whether it is still accurate and fair, without turning your logs into a new data leak.
Prompt and log monitoring
- Prompt monitoring
- Watching inputs and outputs for abuse, injection attempts and data leakage.
- Query
- The prompt content and metadata: user, time, source, size.
- Response
- What the model returned, checked for sensitive data, policy violations and unsafe instructions.
- Log monitoring
- Collecting AI gateway, application and model logs into the SIEM with alerts.
- Log sanitization
- Redacting secrets and personal data from prompts and responses before they are logged.
- Log protection
- Access control, encryption, integrity protection and retention limits for AI logs, which often contain sensitive conversations.
Quality and usage signals
- Response confidence level
- A score of how certain the model is. Low-confidence answers can be flagged for human review or withheld.
- Rate monitoring
- Tracking request rates per user and key to spot scraping, extraction and denial of service.
AI cost monitoring
- Prompts (cost)
- Input tokens consumed; long or repeated prompts drive cost.
- Storage (cost)
- Vector stores, logs, datasets and model artefacts.
- Response (cost)
- Output tokens generated; unbounded outputs inflate cost.
- Processing (cost)
- Compute and GPU time for inference, fine-tuning and embeddings. Sudden spikes can signal abuse (denial of wallet).
Auditing for quality and compliance
- Hallucinations
- Auditing outputs for confident but false statements.
- Accuracy
- Measuring correctness against ground truth over time to detect drift.
- Bias and fairness
- Checking whether outcomes differ unfairly across groups.
- Access
- Reviewing who used the system, what data they reached and whether access was appropriate.
Check your understanding
Loading quiz…
Original CoreQ Academy material, mapped to objective 2.5 of the publicly available CompTIA SecAI+ CY0-001 exam objectives (Exam objectives V1, document version 1.1). Not affiliated with or endorsed by CompTIA. Contains no exam questions.