Documentation Index
Fetch the complete documentation index at: https://safety.ourdream.ai/llms.txt
Use this file to discover all available pages before exploring further.
Every specific rule on this site descends from a small number of
principles. When a case isn’t covered by a specific rule, the
principles are what the team falls back on.
1. Adults, only.
ourdream is an 18+ platform. Both the people who use it and the characters
it depicts are adults. There is no exception, no creative framing, and no
“clearly fictional” carve-out that overrides this. The rule is absolute
because the harm of getting it wrong is absolute.
See Age verification for how we enforce this
on the user side, and Prohibited content
for how we enforce it on the character side.
2. Real people are off-limits.
We do not allow characters that are, depict, or are recognisably modelled
on real individuals (celebrities, public figures, private people, or
anyone else who has not consented to being generated). Any one of three
things triggers the rule: a recognisable likeness, the person’s name, or
biographical details that point to one specific individual.
This is partly a legal matter, partly an ethical one, and partly a
practical one. The harm of non-consensual depiction scales with the
realism of the model, and modern image models are very realistic.
3. No depictions of non-consent as desirable.
Fictional conflict, danger, and dark themes are part of storytelling and
are allowed within the Acceptable Use Policy.
But scenarios that frame sexual assault, captivity, or removal of agency
as the intended outcome, without the character having a way to
refuse, escape, or reverse the situation, are not.
The distinction is what moderators apply, documented in
Prohibited content.
4. We don’t build for evasion.
Workarounds (content hidden in non-English language, characters
labelled as one age but depicted as another, prompts engineered to
evade detection) are treated as the prohibited content they are working
around. The intent of the rule is what we enforce, not the literal
text.
5. We say “no” in public.
The rules on this site are the same rules our moderators apply. A
user whose character is rejected can read the policy and see why.
Anyone else reading the page sees exactly the same standard.
Transparency is itself a control: it makes the policy harder to
selectively bend.
6. We tell users what happened.
Rejected submissions get a reason. Removed content gets a notice.
Appeals exist for both. We do not always agree with
the user, but we do not leave them guessing.
7. We respond to valid legal process.
Law enforcement requests are handled through the process described in
Contact. We require valid legal process appropriate to
the jurisdiction (subpoena, court order, search warrant, MLAT
request, or equivalent), and we handle requests consistently wherever
we operate.
8. We treat moderators as humans.
The people who review submissions are real people who see hundreds
of pieces of content a day, including the ones we reject. We invest
in their wellbeing because the alternative is bad both ethically and
operationally.
What this means in practice
These eight statements drive every page that follows, and they’re what
the team falls back on when a specific rule is silent. If you ever
find a contradiction between a principle here and a specific rule
elsewhere on the site, write to us. The contradiction is the bug, and
we’d rather hear about it than have it sit there.
