I think transparency around training data plays a major role in building trust. As a creator, I'm much more comfortable using AI tools when I know my work isn't being used for model training without my knowledge or consent.

An ethical approach doesn't necessarily make the AI more powerful, but it does make me more willing to integrate it into my workflow. Features like Harmonize become easier to adopt, when the company is clear about where the training data comes from and how user content is handled. Trust and transparency are becoming just as important as the technology itself.

It’s a large trust issue. Not specific to AI, but any company can post anything online about how they are ethically operating. The real question is, what are they actually doing when the only accountability is themselves that their core. AI companies *should* for transparency provide the data sets they are training on for independent review to allow for verification of those claims. My concern is that it’s a large black box and there is no effective way to audit so it’s easy for everyone to say they are ethical.

The question is for users, if an AI company is caught not operating to their code of ethics, are enough of users going to quit using that product to re-enforce the public outcry for data protection, or is the convenience of the tool, and pain of switching going allow said company to avoid consequences? If it‘s the former, then these conversations are largely moot.

D Fosse below mentions model collapse which also sparks me to ask the question: If an “ethical” AI is training on previous non ethical AI creation/outputs, isn’t that kind of a “money washing” scenario for AI training? Would we accept that? It becomes a real gray area of what that ethical boundary is for me. What is the curation process to create that training data set?

How does this effect how I use AI? As a user, it’s kind of just try my best to be good and use good AI - but there is no real way for these tools to be reasonably checked right now. A company to take time to put together a really good ethics policy is certainly better than nothing. 

A much more immediate problem is “model collapse” (look it up) caused by AI cannibalization. You don’t have to scroll down very far in the Stock forum to see that Stock is full of AI generated material - which then is used for training. The result is a steady downwards spiral converging on mediocrity and sameness. Clearly, the filtering isn’t very effective, and Adobe should take a much more aggressive approach to this.

Adobe has made huge contributions to the creative community, for which I’m eternally grateful, but IMO this isn’t one of them. Personally I have never touched the AI tools, and I’m not tempted. I’m not putting my name on something I didn’t make, and that’s the end of it as far as I’m concerned. So I’m not part of this discussion. But I think a little sobriety goes a long way.

Once again, I want to remind everyone that fully 50% of the entire internet is now AI generated. Is this what we want?

I think transparency absolutely affects trust. As someone who works with audio tools, I'm much more comfortable adopting AI features when companies clearly explain what data is used for training and what isn't. Knowing that my projects won't be used without consent removes a major concern.

For example, I often use AI-powered tools for tasks like background music remover and noise reduction. The technology can save hours of manual editing, but I'd be hesitant to upload client recordings if I wasn't confident about how that data would be handled. Clear policies make it easier to focus on creativity instead of worrying about ownership and privacy.

In my experience, ethical AI doesn't necessarily make the output better, but it does make me more willing to integrate it into my workflow. When creators trust the tool, they're more likely to experiment with it and use it as a genuine creative partner rather than just a convenience feature.

AI excels at some things like writing code but falls way short at others that require spatial awareness. 

The Upside Down Cup Test

#ChatGPT #Grok #Gemini #Hume #Claude

Use AI responsibly. It’s not a replacement for human talents. At least, not yet.  And giving AI autonomous control over strategic defense systems is still a pipe dream. We’re nowhere near ready for that.  

This is actually one of the most important conversations around AI right now, especially for creatives.

For me, yes—ethics and transparency directly affect trust. Knowing that Adobe Firefly is trained only on licensed and public-domain content, and not on my personal projects, makes a real difference. It removes that lingering concern of “am I feeding the system my own work without realizing it?”

That’s why features like Harmonize in Adobe Photoshop feel easier to adopt. When the consent boundaries are clear, I can focus on creativity instead of second-guessing how my data might be used later. Transparency doesn’t just build trust—it lowers friction in the creative process.

By contrast, when other AI tools are vague about training sources, it creates hesitation. Even if the results are impressive, uncertainty around data usage can make the tool feel less like a partner and more like a risk.

More broadly, how an AI model is trained absolutely affects how I relate to it creatively. Ethical training and clear communication help position AI as:

• an assistant, not a replacement

• a tool that supports ideas, not extracts value from them

When those principles are in place, I’m far more willing to experiment, iterate, and integrate AI into my workflow.

So yes—responsible data practices don’t just shape trust, they actively shape how and whether AI becomes part of the creative process.

That’s exactly where my question diverges.

I understand—and actually appreciate—Adobe’s stance that Firefly is trained only on licensed and public-domain data, and that it explicitly does not learn from our personal projects. Ethically, that clarity matters.

But creatively, it raises a different issue for me:

How do I get Firefly to work primarily from my own visual language—my imagery, my archive, my authorship—without muddying ownership or consent?

In other words, I’m less concerned about Adobe training on my work, and more interested in deliberately training or steering the model with my work.

Absolutely! Ethical training data is crucial for professional trust. I only use AI-powered creative tools with transparent data practices to protect my work and clients. Transparency isn't optional—it's essential for any serious creative workflow optimization.

I am a big advocate for Ethics in AI.
Which is why I never use Midjourney.
 

I’ve seen far too much stolen copyrighted content in Adobe’s stock images, as well as AI images in Adobe stock that look like they’re from other external AI models which do train on unlicensed data. I appreciate that Adobe is attempting to be more ethical here, but they don’t police their stock library enough to ensure it isn’t full of scammers. Ethically built AI that actually compensates the people it’s built off of and keeps what I make with it private and in my control is the only AI I’d consider using, but Firefly isn’t there yet.