AAISM: Advanced in AI Security Management

By Isaca · Question Mode

QuestionQ11

AI Risk Management

An organization is deploying AI agents that have tools and data access. Which of the following is MOST important to include in end-user acceptable-use policies to reduce misuse?

Explanation

Acceptable-use policies should prohibit prompts intended to manipulate, override, or circumvent an agent’s intended behavior and safeguards. Such direct prompting can be used to bypass output restrictions, manipulate tool use or API calls, and enable harmful actions; NIST identifies these as prompt-injection and jailbreak-related risks.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!