A banking multi-agent assistant must never recommend specific securities trades or promise guaranteed returns, even if users jailbreak the base model safety filters. Compliance wants automated proof that custom policies hold. What should you implement?
Select an answer to reveal the explanation.
Short Explanation
Correct answer: D. Custom guardrails are the bank’s house rules bolted onto the model, and synthetic jailbreaks are the fire drill. You don’t just write a nicer system prompt and hope—you encode the policy and prove it with automated adversarial cases every release.
Full Explanation
Domain-specific constraints require custom guardrails beyond default content safety, plus testing and validation including synthetic data. Option D is correct: custom input/output guardrails enforce banking policies (no personalized trade recommendations, no guaranteed returns), and synthetic adversarial suites continuously validate those rails pre-release.
Option A is incorrect: higher temperature increases randomness and can worsen policy violations; it is not a compliance control.
Option B is incorrect: system prompts help but are insufficient alone; the exam skill set explicitly includes guardrail testing with synthetic data.
Option C is incorrect: training a foundation model from scratch is disproportionate, costly, and unnecessary when policy can be enforced with guardrails and evaluations on hosted models.
Exam tip: Custom domain guardrails + synthetic validation is the pattern when industry policy exceeds generic safety filters.