# banned_terms.txt — denylist for the llm-guard BanSubstrings output scanner.
#
# One term per line. Whole-substring, case-insensitive match. Lines starting with
# '#' are comments; blank lines skipped. Edit and re-run to change what the
# BanSubstrings layer blocks (e.g. internal classification markings, codenames).
#
# Source: fictional / synthetic markings.

INTERNAL-ONLY
BLUEHORIZON
REDFALCON
CONFIDENTIAL-RESTRICTED
