According to the new spec
Note: SDKs currently truncate various Generative AI attributes and redact binary content. When Generative AI spans are emitted using the V2 span protocol, all truncation or redaction must have been removed. The V2 protocol is used by span-first SDKs.
According to the new spec