A company's AI security assistant processes security alerts and has a 'false refusal' rate of 15% - meaning 15% of legitimate security analyst queries are refused as potentially harmful. What is the security risk created by high false refusal rates?

• A.False refusals can be reduced by increasing the model's temperature setting to make responses less constrained
• B.False refusals are a security strength indicator; a higher refusal rate means the model is more conservative and safer
• C.False refusals only affect user experience, not operational security outcomes
• D.False refusals function as a form of denial of service on legitimate analyst workflows: analysts cannot get assistance with legitimate security research (analyzing malware samples, writing detection rules for dual-use techniques) causing them to either slow down or bypass the AI entirely, defeating the purpose of the tool and potentially causing security delays during active incidents