Consider a 2-of-3 multisig address. You can sign a message using any of the 3 keys. But the 2-of-3 address itself is the hash of a script demanding proof from a combination of keys.

So to “sign” a multisig address (P2SH, bc1q…), you have to reveal the script, which reveals the pubkey hashes, then you can sign the message with a quorum of signers.

One could theoretically design a UX for this. Two UXs really—one for the signer and one for the verifier. But to my knowledge, no smooth UX currently exists.