Compliance

A single agent logging PII to an unapproved region or making binding legal statements without audit trails can trigger fines, contract breaches, and loss of customer trust.

Airline safety checklists: every flight follows the same inspected procedures whether the pilot is tired or the weather is clear — compliance isn't optional polish.

Scenario

A healthcare agent detects PHI in a user message, redacts it before the LLM call, routes inference through a BAA-covered endpoint, and stores only hashed session IDs with a 30-day retention policy.

Solution

In Enterprise AI, apply Compliance to this scenario: A healthcare agent detects PHI in a user message, redacts it before the LLM call, routes inference through a BAA-covered endpoint, and stores only hashed session IDs with a 30-day retention policy. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to Compliance (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where Compliance happens in the code.

1# Compliance — minimal example2from openai import OpenAI3
4client = OpenAI()  # create API client5
6# Ask the model to explain this topic7response = client.chat.completions.create(  # core API call for Compliance8    model="gpt-4o-mini",9    messages=[10        {"role": "system", "content": "You explain compliance clearly."},11        {"role": "user", "content": f"What is compliance?"},12    ],13    temperature=0,14)15print(response.choices[0].message.content)  # show output for debugging

• Treating Compliance as a black box without evaluation
• Ignoring cost and latency in production
• Skipping error handling for compliance

Quick recap — the most important points from this module.