OpenAI API Platform
Connect OpenAI API Platform to Safeguard: what it provides, how to set it up, and the frameworks it provides evidence for.
Connect OpenAI API Platform 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 |
| OpenAI API Platform's guide | developers.openai.com |
Connect with an API token
-
In OpenAI API Platform:
- Sign in to platform.openai.com.
- Open the organization's Admin keys settings page (platform.openai.com/settings/organization/admin-keys).
- Create a new admin key and copy its value.
OpenAI API Platform's public documentation does not cover every step.
OpenAI's developer docs link to the Admin keys settings page but do not give the button labels; the help-center article that describes them blocks automated reads (403), so button and permission labels are left out. Only organization Owners can create Admin keys. OpenAI's help center indicates permissions can be chosen at creation; if a read-only set is offered, pick it. Whether the key is shown only once is not confirmed.
Source: developers.openai.com, developers.openai.com, help.openai.com, checked 2026-10-07.
The permissions Safeguard needs:
Admin API key (organization Owner). OpenAI API Platform issues no read-only credential. Safeguard only reads with it, and you can revoke it at any time. Needs an OpenAI Admin API key (sk-admin-…), created by an organization Owner at platform.openai.com › Settings › Organization › Admin keys. A project API key or a ChatGPT Compliance API key does not work here. -
On Compliance › Connectors, open OpenAI API Platform, click Connect, choose API token and fill in:
Field What to enter Organization admin API key Required. An Admin key (starts with sk-admin-), created by an organization Owner at platform.openai.com > Settings > Organization > Admin keys. Stored encrypted and never shown again. -
Click Connect with token.
Frameworks
Safeguard's automated tests read OpenAI API Platform 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 | 12 |
| ISO/IEC 27001 | Automated tests | A.5.16, A.5.17, A.5.18, A.5.34, A.6.5, A.8.2, A.8.5, A.8.10, A.8.15 | 12 |
| HIPAA Security Rule | Automated tests | 164.308(a)(3), 164.308(a)(4), 164.308(a)(5), 164.312(b), 164.312(d) | 8 |
| PCI DSS | Automated tests | 7.2.2, 7.2.4, 8.2.5, 8.4.2, 8.6.3, 10.2.1 | 8 |
| NIST CSF 2.0 | Automated tests | DE.CM-09, GV.OC-03, PR.AA-01, PR.AA-03, PR.AA-05 | 12 |
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