Please adopt c2pa spec. It's a significant step to ensure transparency in genAI
Journey Bunny
I hardly need to mention that there's a lot of feelings about ai-generated content, and LL has already heard quite a lot of feedback about it from residents.
I'd like to suggest a thing we can do about it!
Let me first admit that no solution is
perfect
and that an AI-using-creator could compile, train, and operate AI indpendently without the tech I mention below. But, the 99% use-case is that people generate content using ready-made solutions, and an enormous number of these ai-generating business have committed to a measure of transparency and accountability that I want to mention.If a person generates AI content--marketplace images, for example--using Midjourney, OpenAI (chatGPT), NanoBanana (Google Gemini), Firefly (Adobe), Runaway AI... or any of the myriad next-tier platforms implementing stable diffusion 3, then ALL of these tools automatically embed a content credential into the work. Cropping, resizing, or trivial modifications to the work do not alter the credential. Only significant human-made alterations to the content of the image result in the credential being changed.
The presense of a content credential does not, I should note, mean "made with AI". It means "there's a content credential here." It's important because all major genAI tools
do
generate a credential that describes the origin of the work as a GenAI tool. The content credential carries a modification history with it. This means that even the creator of an original, handmade work could embed a Credential, and if their work was taken and modified by AI, then the credential would list out both versions.Again, all major genAI tools use this. The credentials are already there today. The credentials are not a value judgement, and they are not an ethical judgement. They are already within the ai-generated content, and they help us see what was done with a piece of content. The customer/end user makes the value judgement for themselves.
LL has seen that customers want to know. The information is already within the file. The platform just needs to expose that information. A simple "CR" tag (using the established logo made with the c2pa standard) on the corner of any image on the marketplace, with a hover-tip to explain the content credential would go a long way. Further, adding a property for texture assets generated for use within SL itself would be a nice next-step (but less trivial to implement).
For more info about the industry-wide standard C2PA (Coalition for Content Protection and Authenticity): https://c2pa.org/
The specification is avaialble here:
This isn't a fun, easy feature but it feels like an important feature.
Log In
Moo Boo
Tragically this doesn't apply to code generated by AI which for whatever reason's completely off the chart of controversies (remember: it's only a crime if it's visible!), however I still fully support and condone something like this.
There needs to be a meta tag of sorts for AI generated content.
As of now LL's liberal "let's see if we can provide the AI content before our creators do" attitude is doing nothing but spiraling the community into more confusion and disorientation of which approach to have to things, as well as pinning events against each other due to the whole "who's allowing what" debate.
By now there's events that are basically all AI content, and there's events claiming to have "No AI Allowed" (i'm sure nobody lies when attending those), heck I've even seen an event claiming no-AI only to use AI themselves to make the event (it was very obvious).. It's a mess!
Linden Labs need to start taking responsibility for it's platform in regards to AI content, even if it's just the bare minimum of including a header tag for AI files.
DarkNess Bloodstone
Very important. ⬆
Beatrice Voxel
It IS an important feature.
For those that use AI to generate a 'rough draft' of their work, or alter something they've already created to a new rigging, it would provide details on that workflow (and a creator can then point to the content cred and say "this is where the AI component was used, this is where I tweaked it further".
For those that use AI to copybot other works, the original works (via their creds) would be so flagged as the 'source' for the AI, and AI-powered piracy could indeed be flagged and stopped - as of right now there is zero recourse for ANY kind of copybotting (AI or not), because original works do not have content credentials embedded all that often.
Standard best practices only work when they are used, and it looks like the GenAI companies are trying to establish a standard best practice. SL should participate.