Which technical property of current voice-cloning tools makes them dangerous even when only a 10-second audio sample of the target is publicly available?

• A.Voice cloning requires at least 60 minutes of high-quality audio, so 10 seconds is insufficient for a convincing clone
• B.Voice characteristics are protected by copyright law, preventing commercial use of cloned voices; disinformation campaigns are bounded by generation cost, since producing persuasive long-form content still prices out all but state actors
• C.Voice cloning tools require the target's consent and a signed API agreement before generating audio
• D.Modern voice-synthesis models can learn a speaker's tonal, pitch, and cadence characteristics from a very small sample and generate arbitrary speech in that voice