ChatGPT
Connect ChatGPT to Safeguard: what it provides, how to set it up, and the frameworks it provides evidence for.
Connect ChatGPT to collect business applications evidence with automated compliance tests, switched on per workspace.
| Category | Business Applications |
| Connect with | API token |
| What a connection produces | Automated compliance tests |
| ChatGPT's guide | help.openai.com |
Connect with an API token
-
In ChatGPT:
- Get an Enterprise Compliance API key for your ChatGPT Enterprise or Edu workspace.
Where to find the other values:
- Workspace ID: OpenAI's Compliance API docs say requests use your ChatGPT workspace ID; the exact Admin Console screen showing it could not be read.
ChatGPT keeps the full instructions behind sign-in; this is what its public pages say.
Needs ChatGPT Enterprise or Edu. OpenAI's help-center articles on the Compliance API and on managing Admin keys in the Admin Console block automated reads (403), and the cookbook page only lists the Enterprise Compliance API key as a prerequisite, so who creates the key, the menu path, button labels and permission names are unverified. Confirm in your Admin Console.
Source: help.openai.com, help.openai.com, developers.openai.com, checked 2026-10-07.
The permissions Safeguard needs:
Compliance API. Needs ChatGPT's Enterprise or Edu plan. Needs a ChatGPT Enterprise or Edu Compliance API key and the workspace ID. An API key from platform.openai.com does not work here: connect it with the OpenAI API Platform card. -
On Compliance › Connectors, open ChatGPT, click Connect, choose API token and fill in:
Field What to enter Workspace ID Required. Your ChatGPT workspace's ID: a UUID such as 1a2b3c4d-1a2b-1a2b-1a2b-1a2b3c4d5e6f, shown in the ChatGPT admin console. Not the OpenAI organization ID (org-…). See https://help.openai.com/en/articles/9261474-compliance-api-for-enterprise-customers Compliance API key Required. The Compliance API key OpenAI issues for your ChatGPT Enterprise or Edu workspace. A platform.openai.com key (sk-admin-, sk-proj-) does not work here. Stored encrypted and never shown again. -
Click Connect with token.
Frameworks
Safeguard's automated tests read ChatGPT and map to the controls below. When a test passes, the result is evidence toward each control it maps to, in every framework listed. It is evidence toward a control, not proof that the control is met on its own.
| Framework | Evidence | Controls | Tests |
|---|---|---|---|
| SOC2 | Automated tests | C1.2, CC6.1, CC6.2, CC6.3, CC6.5, CC7.2 | 6 |
| ISO/IEC 27001 | Automated tests | A.5.16, A.5.18, A.5.34, A.6.5, A.8.2, A.8.5, A.8.10, A.8.15 | 6 |
| HIPAA Security Rule | Automated tests | 164.308(a)(3), 164.308(a)(4), 164.312(b), 164.312(d) | 5 |
| PCI DSS | Automated tests | 7.2.2, 7.2.4, 8.2.5, 8.4.2, 10.2.1 | 5 |
| NIST CSF 2.0 | Automated tests | DE.CM-09, GV.OC-03, PR.AA-01, PR.AA-03, PR.AA-05 | 6 |
These tests are switched on per workspace by the Safeguard team.
Related
- Business Applications connectors
- Compliance connectors: how connecting works, card statuses and multiple accounts
- Connector setup reference: every connector in one table
- Frameworks and connectors: how connectors provide evidence for frameworks