COMPARISON
Locke vs Internal AI Portals and Private AI Deployments
Many organisations build an internal AI portal — a private ChatGPT-like interface backed by Azure OpenAI, Bedrock, or a similar managed endpoint, with SSO, role-based access, audit logging, and vetted terms. This is sound governance: it centralises AI access, creates an audit trail, and cuts down on shadow AI. The gap is content. It governs the channel and the identity, not what gets typed. Employees can still paste sensitive data into the approved prompt box, and the portal has no way to catch it at the field level.
When Internal AI portals and private AI deployments fits
- You need centralised governance: a single approved interface, SSO enforcement, usage logging, and model version control.
- You need to prevent shadow AI by giving employees a sanctioned alternative to consumer accounts.
- You need an audit trail of who used AI, when, and what model — at the organisational level.
When Locke fits
- Even with a vetted portal, sensitive data can still appear in prompts — the portal controls the channel, not the content.
- You want to prevent specific categories of sensitive data from reaching any model, including your own internal portal.
- You need a control that works at the moment of composition, before text is submitted to the portal backend.
Side-by-side
| Dimension | Locke | Internal AI portals |
|---|---|---|
| What it governs | The content of prompts — which sensitive fields appear in submissions. | The channel — who uses AI, through which interface, with what model. |
| Can it prevent PII from entering a prompt? | Yes — detects and masks sensitive data before the prompt is submitted. | No — inspects requests at the API level after the user submits. |
| Audit trail | Local detection log of what was flagged and how it was handled. | Full audit log of requests, users, timestamps, and model responses. |
| Shadow AI risk | Controls content on approved tools; doesn't prevent access to unapproved ones. | Reduces shadow AI by offering a sanctioned alternative; doesn't eliminate it. |
| Best for | Preventing sensitive data from reaching any model, including approved portals. | Centralising governance, enforcing approved tooling, and creating audit records. |
Bottom line
Internal AI portals and Locke are complementary. The portal tells you who used AI, through an approved, audited channel. Locke makes sure that even through that channel, the most sensitive fields never reach the model. Together, they cover both the governance layer and the data layer.
