Review burden is the amount of human checking required to validate AI output before it can be trusted or used. In cyber security, it measures how much analyst time is spent verifying alerts, triaging suspicious messages, checking code, or correcting AI-generated guidance. A lower review burden can mean faster operations, but only if quality stays high and errors do not increase.
This term matters because attackers and defenders both use AI in high-volume workflows. Security teams may use AI to summarize logs, rank threats, or draft incident notes, while attackers may use it to generate phishing text or automate reconnaissance. If the review burden stays high, AI is not reducing risk or workload much. If it falls safely, it can show that the system is becoming more reliable and easier to govern.