Which term describes the practice of deliberately trying to find flaws, vulnerabilities, and failure modes by attacking a system?

Which term describes the practice of deliberately trying to find flaws, vulnerabilities, and failure modes by attacking a system?

• A. Red-Teaming
• B. AI Safety
• C. AI Alignment
• D. Frontier Model

Answer: (A)

Red-teaming is the practice of deliberately trying to find flaws, vulnerabilities, and failure modes by attacking a system. It involves simulating an attacker’s perspective to probe defenses, logic, and resilience, uncovering weaknesses that defenders might not anticipate. The goal is to reveal gaps so they can be fixed before real harm occurs, strengthening safety and reliability. This differs from broader aims like AI Safety or AI Alignment, which focus on ensuring AI behaves safely and in line with human values rather than on attacker-style testing. Frontier Model refers to the latest-generation models themselves, not a testing method, so it doesn’t capture the described activity.

Red-teaming is the practice of deliberately trying to find flaws, vulnerabilities, and failure modes by attacking a system. It involves simulating an attacker’s perspective to probe defenses, logic, and resilience, uncovering weaknesses that defenders might not anticipate. The goal is to reveal gaps so they can be fixed before real harm occurs, strengthening safety and reliability. This differs from broader aims like AI Safety or AI Alignment, which focus on ensuring AI behaves safely and in line with human values rather than on attacker-style testing. Frontier Model refers to the latest-generation models themselves, not a testing method, so it doesn’t capture the described activity.