Answered

User guidelines AI needs a big tweak

Forum|Forum|4 months ago
January 7, 2026
8 replies
1026 views

What is going on with the "user guidelines" saying my children's storybook pictures (example of offending image attached) are not acceptable? They were created using Firefly! How could it be unacceptable to use those images to make a video? Please retrain the AI to understand this.

Firefly_Gemini Flash_Have the boy assistant be assist 959858 P7u.png

Generate Video

Correct answer droopydog500

Hi @KeenEyeD,

[I had to edit me response because the forum software said my response contained se[x]ual content.]

The reasons are rooted in profound ethical, legal, and safety considerations. Adobe has not explain the reasons why, but in general, multiple AI companies have done this for these reasons:

1. Prevention of Child Se[x]ual Abuse Material
The most critical reason is the prevention of the creation or augmentation of these materials. AI video generators could be exploited to create non-c[o]nsensual, se[x]ually e[x]plicit, or suggestive content involving minors. By blocking the processing of images containing children, companies create a "hard barrier" against the production of illegal and predatory material.

2. Risk of "Deepfake" E[x]ploitation and Harassment
Even if the generated content is not e[x]plicitly se[x]ual, AI can be used to place children in dangerous, h[u]miliating, or compromising situations. These "deepfakes" can be used for cyber-bullying, e[x]tortion, or to damage the reputation of a minor before they are old enough to manage their own digital footprint.

3. Lack of Informed Consent
Minors cannot legally provide informed consent for their likeness to be used, transformed, or distributed by AI models. Since the AI processes and "re-imagines" the source image, it creates a new derivative work. Tech companies argue that animating a child's likeness without the ability for that child to understand the long-term implications is an ethical violation.

These are the most likely reasons why this restriction exists.

droopy

daniellei4510

Community Expert

Another anomaly. Nana Banana (not the Partner model) refused to generate an image of a baby swaddled in a blanket. As though it interpreted "swaddled" as "smothered." Weirder yet, while most AI models refuse to generate images or even edits when too much skin is visible, I've had two or three examples of completely nude models generated without even coming close to asking for same. It apparently depends on how the AI INTERPRETS the prompt, rather than what it actually states.

Adobe Community Expert | If you aren't submitting your assets in sRGB, you probably didn't read the rules.

daniellei4510

Community Expert

As I mentioned elsewhere in this post, this issue of censorship is not restriced firefly alone. I just tried to create an image outside of Firefly. on Nano Banana of a young girl riding a giraffe and it refused to create it, responding with: "I can create images for you, but not ones that depict minors like that. Can I help with a different image instead?" It's an illogical restriction, based on how so-called "intelligent" the so called intelligence in AI actually is.

Adobe Community Expert | If you aren't submitting your assets in sRGB, you probably didn't read the rules.

K

KeenEyeDAuthor

Reminds me of a bumper sticker highlighting that sometimes our safety
precautions are a little unbalanced..

droopydog500

Correct answer

Community Manager

Hi @KeenEyeD,

[I had to edit me response because the forum software said my response contained se[x]ual content.]

The reasons are rooted in profound ethical, legal, and safety considerations. Adobe has not explain the reasons why, but in general, multiple AI companies have done this for these reasons:

1. Prevention of Child Se[x]ual Abuse Material
The most critical reason is the prevention of the creation or augmentation of these materials. AI video generators could be exploited to create non-c[o]nsensual, se[x]ually e[x]plicit, or suggestive content involving minors. By blocking the processing of images containing children, companies create a "hard barrier" against the production of illegal and predatory material.

2. Risk of "Deepfake" E[x]ploitation and Harassment
Even if the generated content is not e[x]plicitly se[x]ual, AI can be used to place children in dangerous, h[u]miliating, or compromising situations. These "deepfakes" can be used for cyber-bullying, e[x]tortion, or to damage the reputation of a minor before they are old enough to manage their own digital footprint.

3. Lack of Informed Consent
Minors cannot legally provide informed consent for their likeness to be used, transformed, or distributed by AI models. Since the AI processes and "re-imagines" the source image, it creates a new derivative work. Tech companies argue that animating a child's likeness without the ability for that child to understand the long-term implications is an ethical violation.

These are the most likely reasons why this restriction exists.

droopy

Adobe Community Expert (not an Adobe employee)

K

KeenEyeDAuthor

Wow, droopy! You went to a lot of effort there. I am most impressed by your
workaround to submit the article! Well done, thanks for persevering!
I'm sure most creatives who use Adobe products and other generative AI
resources have no intention of using the tools for nefarious ends. It is so
incredibly perplexing to someone without these evil intents, who just wants
to create something healthy for students' educations. To get a "you're a
stupid monkey for even thinking we were gonna let you create a video using
images that go against our minority report styled policy", after the images
themselves were created using the same product, well, let's just say, that
does something to my pancreas that this creative can't put into words..
It actually hurts my heart to see my attempts to create be shot down in
flames like this, makes me feel like I'm being judged, and makes me nervous
for what more roadblocks are going to present themselves on my journey
trying to achieve accurate output in a productive way from these tools.
All the best with your journeys..

droopydog500

Community Manager

hi @KeenEyeD,

You can generate images that have children in them. What is blocked is using an image with children as a keyframe for a video generation.

droopy

Adobe Community Expert (not an Adobe employee)

K

KeenEyeDAuthor

Explain to me the sense in that.

daniellei4510

Community Expert

Another example: I once had a prompt that, among other thinks, prompted for an image of "two beautiful twin girls." This is what the AI returned:

Two Cats.jpg

Adobe Community Expert | If you aren't submitting your assets in sRGB, you probably didn't read the rules.

daniellei4510

Community Expert

Yes, it's frustrating, but creators are at the mercy of the AI and how it's programmed, apparently. I once prompted in Gemini for a scene of a little girl screaming as she rode down a slide, but the bot came back with the message that it will not create images that indicate danger to a child. Or something to that effect. I had to change "screaming" to "laughing with her mouth open" or something along those lines. It comes down to outsmarting the "intelligence" part of artificial intelligence.

Adobe Community Expert | If you aren't submitting your assets in sRGB, you probably didn't read the rules.

O

Oh.N8

@KeenEyeD

Just to add to what @daniellei4510 mentioned... Adobe and other model are not going to generate any images with children. It is frustrating for artists like yourself 100%. The guidelines are there for a reason.

Cheers

Nate

K

KeenEyeDAuthor

What do you mean? I have a stack of images, all generated in Firefly, that
have cartoon characters of children in them. I'm sure if you go and chuck
in some prompts, you'll have children output sooner than you can say
'schooled'.

daniellei4510

Community Expert

Firefly (along with other AI models within and outside of Firefly) do not accept images of children to create videos.

Adobe Community Expert | If you aren't submitting your assets in sRGB, you probably didn't read the rules.

K

KeenEyeDAuthor

Actually, there are a bunch of similar scenes with children in them and
they were not flagged. I believe, after further trials, it was CHILDREN +
FIRE that tripped the system, despite the story being about a workplace
hazard and the output essentially teaching fire safety. The input is
cartoon drawings and there is no actual depiction of violence or gore.
However the threat of danger presented in the fire in the same scene as
children present, seems to be the flagging criteria. I've just used a later
picture in the sequence to create a video. The fire is almost out and the
risk of danger is over. Children are present in the scene. The video
worked. Go figure.
How can AI be given clearance to use 'usual threats of danger' that
children need to be educated in for their safety? And if the material is to
be inclusive for them, that means it should contain CHILDREN + FIRE
elements together on the same page.
Nobody gets hurt, wounded or worse in the story, but of course this is the
narrow pointy end of the stick, held to account on a single heavy-handed
policy that is shooting itself in the foot.

Sign up

To post, reply, or follow discussions, please sign in with your Adobe ID.

Sign in to Adobe Community

To post, reply, or follow discussions, please sign in with your Adobe ID.

Scanning file for viruses.

This file cannot be downloaded