OpenAI ChatGPT Enterprise Custom Agent Published


Description

Detects ChatGPT Enterprise custom-agent publish events (agent_published). Publishing an agent with connectors, skills, or store visibility can establish durable automation and broaden data access beyond the original user.

Query · esql

from logs-openai_chatgpt_enterprise.custom_agents_log-* metadata _id, _version, _index
| where
    data_stream.dataset == "openai_chatgpt_enterprise.custom_agents_log" and
    event.action == "agent_published"
| keep
    _id, _version, _index, @timestamp, event.action, user.id, user.email,
    gen_ai.agent.id,
    openai_chatgpt_enterprise.custom_agents_log.client_id,
    openai_chatgpt_enterprise.custom_agents_log.event_details.publish_type,
    openai_chatgpt_enterprise.custom_agents_log.event_details.release_name,
    openai_chatgpt_enterprise.custom_agents_log.event_details.agent_fields.store_visible,
    openai_chatgpt_enterprise.custom_agents_log.event_details.agent_fields.workspace_capability,
    data_stream.*

Investigation fields

Pivot points the source recommends for triage.

  • @timestamp
  • event.action
  • user.id
  • gen_ai.agent.id
  • openai_chatgpt_enterprise.custom_agents_log.event_details.agent_fields.store_visible
  • openai_chatgpt_enterprise.custom_agents_log.event_details.publish_type

Implementation guide

This rule requires the OpenAI ChatGPT Enterprise integration to collect compliance logs into Elasticsearch.

https://www.elastic.co/docs/reference/integrations/openai_chatgpt_enterprise

Known false positives

  • Platform teams publish approved custom agents during workspace rollouts.

Analyst notes

Investigating OpenAI ChatGPT Enterprise Custom Agent Published

A custom agent was published. Published agents with connectors, skills, and store visibility can establish durable automation and broaden data access.

Escalate when visibility is broader than expected, connectors are sensitive, or the publisher is not authorized to release agents. Close as a false positive for approved platform rollouts.

Possible investigation steps

  • Identify the actor and agent identifiers on the alert.
  • Review connectors, skills, store visibility, and any release notes or description fields.
  • Confirm whether publication matches an approved change request.
  • Look for related trigger creation or OAuth link events from the same actor.

False positive analysis

  • Platform teams publish approved custom agents during controlled workspace rollouts.

Response and remediation

  • Unpublish or delete unauthorized agents and revoke connected OAuth links.
  • Review recent agent runs if the agent may already have executed.
Raw source OpenAI ChatGPT Enterprise Custom Agent Published · Elastic TOML
Esc
Published by elastic/detection-rules ↗, licensed under Elastic License 2.0 ↗. Reproduced here unmodified.
[metadata]
creation_date = "2026/10/06"
integration = ["openai_chatgpt_enterprise"]
maturity = "production"
updated_date = "2026/10/06"

[rule]
author = ["Elastic"]
description = """
Detects ChatGPT Enterprise custom-agent publish events (`agent_published`). Publishing an agent with connectors,
skills, or store visibility can establish durable automation and broaden data access beyond the original user.
"""
false_positives = [
    "Platform teams publish approved custom agents during workspace rollouts.",
]
from = "now-9m"
language = "esql"
license = "Elastic License v2"
name = "OpenAI ChatGPT Enterprise Custom Agent Published"
note = """## Triage and analysis

### Investigating OpenAI ChatGPT Enterprise Custom Agent Published

A custom agent was published. Published agents with connectors, skills, and store visibility can establish durable automation and broaden data access.

Escalate when visibility is broader than expected, connectors are sensitive, or the publisher is not authorized to release agents. Close as a false positive for approved platform rollouts.

#### Possible investigation steps

- Identify the actor and agent identifiers on the alert.
- Review connectors, skills, store visibility, and any release notes or description fields.
- Confirm whether publication matches an approved change request.
- Look for related trigger creation or OAuth link events from the same actor.

### False positive analysis

- Platform teams publish approved custom agents during controlled workspace rollouts.

### Response and remediation

- Unpublish or delete unauthorized agents and revoke connected OAuth links.
- Review recent agent runs if the agent may already have executed.
"""
references = [
    "https://www.elastic.co/docs/reference/integrations/openai_chatgpt_enterprise",
    "https://platform.openai.com/docs/compliance",
]
risk_score = 47
rule_id = "c73d406c-4399-40f2-a255-4a42f8ead9e0"
setup = """## Setup

This rule requires the OpenAI ChatGPT Enterprise integration to collect compliance logs into Elasticsearch.

https://www.elastic.co/docs/reference/integrations/openai_chatgpt_enterprise
"""
severity = "medium"
tags = [
    "Domain: GenAI",
    "Platform: OpenAI",
    "Data Source: OpenAI ChatGPT Enterprise",
    "Service: OpenAI ChatGPT Enterprise",
    "Use Case: Threat Detection",
    "Resources: Investigation Guide",
    "Rule Type: ES|QL",
    "Tactic: Persistence",
    "Mitre Atlas: AML.T0081",
]
timestamp_override = "event.ingested"
type = "esql"

query = '''
from logs-openai_chatgpt_enterprise.custom_agents_log-* metadata _id, _version, _index
| where
    data_stream.dataset == "openai_chatgpt_enterprise.custom_agents_log" and
    event.action == "agent_published"
| keep
    _id, _version, _index, @timestamp, event.action, user.id, user.email,
    gen_ai.agent.id,
    openai_chatgpt_enterprise.custom_agents_log.client_id,
    openai_chatgpt_enterprise.custom_agents_log.event_details.publish_type,
    openai_chatgpt_enterprise.custom_agents_log.event_details.release_name,
    openai_chatgpt_enterprise.custom_agents_log.event_details.agent_fields.store_visible,
    openai_chatgpt_enterprise.custom_agents_log.event_details.agent_fields.workspace_capability,
    data_stream.*
'''


[[rule.threat]]
framework = "MITRE ATT&CK"
[[rule.threat.technique]]
id = "T1176"
name = "Software Extensions"
reference = "https://attack.mitre.org/techniques/T1176/"


[rule.threat.tactic]
id = "TA0003"
name = "Persistence"
reference = "https://attack.mitre.org/tactics/TA0003/"

[[rule.threat_mappings]]
framework = "MITRE ATLAS"
version = "2026.08"
[[rule.threat_mappings.threat]]
framework = "MITRE ATLAS"
[[rule.threat_mappings.threat.technique]]
id = "AML.T0081"
name = "Modify AI Agent Configuration"
reference = "https://atlas.mitre.org/techniques/AML.T0081/"


[rule.threat_mappings.threat.tactic]
id = "AML.TA0006"
name = "Persistence"
reference = "https://atlas.mitre.org/tactics/AML.TA0006/"

[rule.investigation_fields]
field_names = [
    "@timestamp",
    "event.action",
    "user.id",
    "gen_ai.agent.id",
    "openai_chatgpt_enterprise.custom_agents_log.event_details.agent_fields.store_visible",
    "openai_chatgpt_enterprise.custom_agents_log.event_details.publish_type",
]

Detection rules belong to the projects that publish them and remain under their own licenses. This site indexes and links to them; it claims no rights in them.