Security wants shift-left testing of a multi-agent Foundry solution for jailbreaks, tool abuse, and data exfiltration attempts before external launch. Which capability should you incorporate into the SDLC?
Select an answer to reveal the explanation.
Short Explanation
Answer C. Shift-left security for multi-agent solutions includes the AI Red Teaming Agent in Foundry. Adversarial probes for jailbreaks, tool misuse, and exfiltration should fail the build or release gate when severity is high. Waiting for breaches, load-only tests, or disabling safety all miss the requirement.
Full Explanation
Correct answer: C. Implement shift-left security principles, including the AI Red Teaming Agent in Foundry. Integrate adversarial scenarios into CI/pre-prod, track findings, and remediate via guardrails, tool permission tightening, and prompt defensive guidelines.
A is incorrect: post-breach-only testing is the opposite of shift-left.
B is incorrect: performance tests do not substitute for adversarial AI security testing.
D is incorrect: removing safety systems increases risk and still needs validation.
Feed red-team cases into the same evaluation suites used for quality regression.