# Trust and safety

Report anything from the flag on a listing page, or write to safety@contextfile.ai.

We act on misrepresented verification reports, sources used without a license,
prompt-injection attempts in connector content (quarantined automatically, then
reviewed by a person), and fraud patterns in the trust graph such as floor-fishing.

Enforcement is accountable too. Every action writes an edge to the same append-only
graph, sellers can appeal, and an upheld appeal adds a correcting edge in public. The
transparency digest publishes enforcement counts by category; we never publish the
detection rules that would teach evasion.



Version 1, updated 2026-10-11.