AI applications tend to aggregate more data into observable context.
IDENTITYAI SECURITY ENGINEERING
Secrets in AI systems
Secrets can leak through prompts, tool output, traces, configuration, repositories and model context.
A tool returns an environment file and the full response is captured in tracing.
Which security control should be explicit?
REMEMBERAssume prompts and traces are production data stores.
No uploads · No company data · No account required