Our AI policy

As some of you may know, I (Rae) have spent over two decades in software development, and have established myself as a machine-learning expert. Machine-learning is a subset of AI, and in modelling, we look very hard on the local and global model quality and associated error/uncertainties. I work closely with colleagues who are developing LLMs for generative chemistry.

Based on this background, I have a lot to say about the statistical irrelevance of AI detection software and how it's a fallacy to blindly assume any of them are even remotely accurate.

I also have a strong belief that information that is proprietary to myself or a business must remain private.

Hear me out.

In the software industry, we provide a finished product. The user NEVER gets to see any of the source code, nor should they. Hacking into source code and reverse engineering both constitute violations of End User License Agreements.

I feel the same way about the creations with which our authors and illustrators have trusted us. We take great care to satisfactorialy establish provenance/rights internally. We look at drafts, layered image files, scribbles on paper, and more.

Outside of the Dapple Grey Press acquisition team, this validation will not be made available. Readers can choose to trust our internal processes, or choose not. Vote with your money.

We believe creators deserve both protection from genuine misuse of artificial intelligence and protection from unsupported accusations about their own work.

Here's the statement of our position:

*****************************************************************

We recognise that concerns about generative AI within creative industries are legitimate. Authors and illustrators are entitled to expect their work not to be appropriated, imitated, or commercially exploited without proper rights or permission.

However, we do not support public accusations based solely on aesthetic judgment, automated detection, perceived stylistic characteristics or online speculation.

Statements such as "this looks AI-generated" are not evidence of provenance, and will be interpreted as defamatory.

Dapple Grey Press will not require its authors, illustrators or other contributors to publicly prove their humanity by releasing proprietary working materials in response to unsupported allegations. Where a credible rights or authorship concern is raised, we will investigate it privately and appropriately.

*****************************************************************

Is this a perfect system? Well, as much as we'd like to think it is, it probably isn't. LLMs are improving nearly daily, and it's inevitable that we will miss something with our eyeball review. But I'd rather miss something than make a false accusation that has no merit beyond vibes and a low-confidence AI model that's grading its own homework.

Our public policy is simple. We will not be drawn into any social media witch-hunts. An accusation made by random internet people based on "vibes" or a "detector" will not be addressed.

We will take a page from the British Royal Family playbook: Never complain, never explain.

However, in cases where accusations are made in a way that could result in reputational and/or material damage to Dapple Grey Press Ltd. and/or our authors/creators, we will engage our money-grubbing lawyers and proceed accordingly.