LoFP LoFP / openai_chatgpt_enterprise

openai_chatgpt_enterprise rule

TitleTags
administrators add legitimate corporate domains during onboarding or m&a.
administrators deleting expired workspace log archives under an approved retention policy.
administrators intentionally enable or disable workspace features during rollout.
approved automation that uses chatgpt's slack connector for triage or knowledge lookup. raise the distinct-query threshold if those roles remain noisy.
approved automation, compliance api tooling, or sso libraries that use non-browser user agents.
approved mcp apps such as slack or ticketing connectors generate routine traffic during normal work. first-seen alerts are common during onboarding when a user first uses an approved app.
approved prompt-security or dlp validation tests may intentionally submit synthetic secrets.
approved red-team or prompt-security evaluations that intentionally elicit refusals. exclude those users or raise the count threshold.
approved rollout of codex plugins such as slack, gmail, or google drive for engineering or support teams. confirm the plugin against your workspace connector inventory and change process before escalating.
cloud egress, split-tunnel, or dual-homed clients that present different public ips for concurrent sessions.
common google drive/gmail and slack connector app_id values are excluded; other sanctioned connectors may still alert.
developers create codex tokens for ci runners and local environments.
developers may paste example keys while debugging or documenting integrations.
discussions of certificate or tls rotation can match loosely worded replies; this rule requires secret-family terms plus exposure or recognition phrasing to reduce those cases.
documentation samples and example keys can match token patterns. confirm the value is live before rotating it.
engineers enable approved codex mcp connectors for github, slack, ticketing, and internal tools during normal development. validate the tool name, connector, and whether the integration matches an approved workflow.
hr, legal, finance, or ir users who legitimately search gmail for attachments, payroll, or contract terms via chatgpt.
hr, legal, finance, or vendor-management workflows that legitimately search drive for payroll, nda, or contract terms across several queries. validate the query values and business context before escalating.
incident responders, support engineers, or community managers who run many slack search refinements in a short window. validate the query values and business context before escalating.
it administrators legitimately promote users to account-owner during staffing changes.
it and hiring workflows routinely invite new members during onboarding.
long helpful answers that incidentally include phrases such as \"i can't help with x, but i can help with y.\" read the sampled messages before treating the alert as a hard refusal campaign.
mobile clients on cellular networks that peer through regional hubs may geo-resolve differently than the user location.
operations teams create approved triggers for recurring agent workflows.
platform teams publish approved custom agents during workspace rollouts.
power users and legitimate multi-connector codex workflows (for example slack + github + google drive in one session) can approach these thresholds. raise `esql.gen_ai_tool_name_count_distinct` or `esql.connector_id_count_distinct` above your baseline before broad enablement.
power users iterating many related document titles in one session. raise the distinct-query or sensitive-query thresholds if approved roles remain noisy.
security and compliance teams download workspace logs for investigations, siem validation, or regulatory requests.
security engineering teams legitimately onboard repositories and tune scan lookback windows.
teams intentionally publish gpts or skills for approved internal or partner use under change management.
users delete memories during privacy cleanup or troubleshooting.
users downloading many attachments during legitimate research or migration work.
users drafting security guidance, runbooks, or support replies that quote rotate or revoke advice without an in-chat exposure. confirm the assistant is reacting to a secret in the conversation.
users exporting or reviewing many drive files through chatgpt during approved research, ediscovery, or migration work.
users intentionally share conversations for collaboration within the organization.
users legitimately asking chatgpt to send mail through an approved gmail connector.
users legitimately connect approved apps during onboarding.
users mistype passwords repeatedly or sso misconfiguration can generate clusters of failed logins.
users on vpn or proxy egress that geo-resolves through a region distant from the user physical location.
users on vpn or proxy egress that geo-resolves through multiple regions in a short window.
users purging old chats during workspace hygiene or retention cleanup.