California passed a new AI labeling bill last month and im seeing wildly different interpretations of what it actually requires. specifically: does it apply to content creators using AI in their workflow, or only to platforms? does it require visible labels or just metadata? trying to understand what im supposed to do as a CA-based consultant before the effective date
Read the bill text twice. The short version: it requires platforms with X+ users to surface labels for synthetic content, and it requires creators using AI in a way that could deceive to disclose. the ‘could deceive’ phrasing is key, it applies to manipulated content presented as authentic, not to using a grammar assist. CA AG will issue clarifying guidance in q3.
plus one to the deception framing. its not ‘used AI in workflow’ that triggers disclosure, its ‘created content that could be mistaken for unmanipulated reality’. that distinction is going to confuse everyone for at least 12 months.
Ok so for my b2b consultancy work this is basically a no-op unless im producing synthetic media. understood
The trick will be the enforcement guidance once it lands. the wording leaves a lot of room for selective enforcement.
from the publishing side: we’ve added the disclosure language to our submission guidelines defensively. doesnt matter if it technically applies to us yet, the cost of compliance is low and it sets the right expectation with contributors.
The q3 enforcement guidance will be the actual signal. anyone who panics before then is overcorrecting. read the bill, hold off, watch the AG.