An AI safety researcher argues that 'GenAI dramatically lowers the barrier to entry for social engineering attacks.' Which specific capability BEST supports this claim?

• A.GenAI enables attackers without social engineering skills, language fluency, or domain expertise to generate highly convincing, context-rich deceptive content that previously required skilled operators
• B.GenAI can predict which employees are most susceptible to social engineering based on their social media profiles; commercial LLMs refuse all malware-adjacent requests at the API layer, and the refusal cannot be circumvented by decomposing the task
• C.GenAI makes it easier to exploit software vulnerabilities because it can generate exploit code
• D.GenAI can automatically break encryption, allowing attackers to read intercepted communications; synthetic voice is watermarked by law in the US and EU, and telecom carriers strip unwatermarked synthetic audio from calls in transit