https://reclaimthenet.org/senators-curb-deceptive-ai-and-deepfakes
There are some things to consider here:

1) Facial recognition is worst when the government is using it, particularly for law enforcement purposes. It can also be bad when private companies are using it for surveillance purposes.

I don't think there needs to be a new special regulator who issues licenses to boot to deal with this though.

2) I'm not sure anyone is using "deepfakes" in election ads for deceptive purposes. Still, if it is simply for campaign ads, that isn't as bad as exaggerated worries about their political impact more broadly on social media.

I'll leave this for free speech people to dig into though.

3) "The proposed mandate will require government licensing for firms developing AI technologies with a high potential for misuse including large language models like those developed by OpenAI"

Every technology has a "potential for misuse". This is not a compelling argument.

I'm not a fan of the government trying to get involved here, especially as a fair bit of discourse around LLMs and "potential harms" has been abysmal. Also, the government not approving of people engaging in particular forms of speech looks a lot like a violation of the First Amendment.

For politics specifically, while someone could abuse this tool (as they could other tools), it hasn't been shown it is that impactful. Even if it had some impact, it doesn't stand to reason that a heavy-handed regimen which would likely go far beyond that (as is often the case with government) is the right course of action.