On October 28, VRChat said it will not add AI-generated labels or filters.
The big change
- What changed: Visitors will lack the proposed platform AI-generated label and filter.
- Why it matters: A world’s appearance or rating cannot establish how it was made.
- What to watch: A voluntary declaration can describe a process, but its absence does not explain why it is missing.
The decision and what comes next
VRChat cites unreliable self-reporting and harassment risks, and says filters depend on accurate labels.
It is considering an optional self-declared “Human-authored” mark and a “Tools used” field, but has promised neither.
World likes and dislikes are targeted for late 2026 or early 2027; the dates are estimates. A rating communicates feedback about an experience, not the tools used to create it.
VRChat reports 21,595 survey responses, with 95% wanting some AI-use label. The voluntary survey was weighted toward its player base; it describes respondents, not all users.
An earlier statement had left labeling and filtering under consideration and prohibited misleading thumbnails whether AI was used or not. A feature request by °sky sought mandatory disclosure and filtering; it reflects that user’s proposal, not a representative poll. BIG CHANGE reviewed the statements and request but did not test VRChat’s features or enforcement.
