I suspect that the amount of effort to craft a convincing deepfake is still fairly substantial.

Not to mention most of the major AI providers have safeguards built in against this kind of thing. I can't get Chat or Grok to generate images of Xi Jinping for example, or of copyrighted content.