AI credentials

Manage API keys and credentials used by the integrated AI providers.

AI Credentials

The AI Credentials screen is where your organisation connects an external AI provider to Priverion DPMS. Until you complete this setup, every AI-assisted feature in the platform — auto-generated descriptions, compliance text suggestions, document summaries, and more — will remain unavailable. Think of this screen as the master switch for AI across the entire application: once you save a valid, active credential here, AI helper buttons light up on ROPA records, vendor questionnaires, asset descriptions, and every other module that supports them. This screen is most commonly visited by IT administrators during initial platform setup, and revisited whenever your organisation changes AI provider, rotates an API key, or wants to temporarily disable AI features.

How to open it

Navigate to IT Settings in the main application sidebar. In the left-hand navigation panel that appears, expand or click the AI section, then select Credentials. The breadcrumb at the top of the page will confirm your location: IT Settings › AI › Credentials.

You need the IT Settings – AI read permission to view this screen. Without it, you will see a standard access-denied page. To edit the configuration, you additionally need the IT Settings – AI edit permission. If you can view the summary but not edit it, the Edit button will redirect you to an access-denied page when clicked — this is expected behaviour for read-only roles such as compliance officers.

What you see

The screen has two modes that share the same URL: a summary view (shown by default) and an edit form (opened from the summary view).

On the summary view, a single card fills the page. Its heading reads AI Credentials and an Edit button sits in the top-right corner of the card. The card body shows the currently saved configuration as labelled rows. If no credentials have been saved yet, the only row reads Provider: No active provider configured. Once credentials exist, you see the provider name, a partially masked API key, the selected default model, the token limits for that model, and a Status row showing either Active or Inactive.

When you click Edit, the card is replaced by a full-width form. A back arrow in the top-left returns you to the summary view without saving. The form stacks all input controls vertically, with a blue accent bar visually grouping related fields. At the bottom, a Save button commits your changes.

Working with this screen

Setting up your AI provider for the first time

When your organisation has just deployed DPMS and no AI credentials have been saved yet, the summary card will display No active provider configured. To set up AI, click Edit.

The first thing to do is choose your provider. Click the AI Provider dropdown and select the AI company whose API your organisation has signed up with — for example, Anthropic, OpenAI, or Google. Once you make a selection, the form expands with the specific fields that provider requires. For Anthropic and OpenAI, this will be an API Key field; for Google, there may be additional options.

Enter your API key by clicking the field labelled API Key and typing or pasting the key from your provider's developer console. The field shows a placeholder prompt until you start typing.

Next, click the Default Model dropdown and choose the model you want DPMS to use for all AI-assisted operations. As soon as you select a model, two pieces of information appear below the dropdown in small grey text: the Context Window (for example, 200,000 tokens) and the Max Output (for example, 8,192 tokens). These figures tell you how much content the AI can read in one request and how long its responses can be, which is useful context before committing to a model.

Once you are satisfied with the provider and model, turn on the Activated toggle so it turns blue. This is the step that makes the integration live — without it, your credentials are saved but AI remains disabled everywhere. Finally, click Save. A success notification will confirm the save, and you will be returned to the summary view, which now shows your chosen provider, masked API key, model name, token limits, and a Status of Active. From this moment, AI helper buttons will become available throughout DPMS.

Updating an existing credential (rotating an API key or switching models)

If your provider has issued a new API key — for example, as part of a security rotation — or if you want to switch to a newer, more capable model, navigate to AI Credentials and click Edit.

The form loads with your existing configuration pre-filled. The API key field shows a masked preview such as sk-***. You do not need to re-enter the key to change only the model. Simply click the Default Model dropdown, select the new model, review the updated token limits that appear beneath it, and click Save. The existing key is preserved because you never clicked the key field.

If you do need to update the API key, click anywhere inside the API Key field. The masked preview clears, and the field switches to a secure input mode. Type or paste your new key, then click Save. The old key is replaced.

Heads up: If you change the AI Provider dropdown at any point, the entire form resets immediately and without warning — your new API key, model selection, and toggle state are all cleared. Select your provider first, then fill in the other fields.

Disabling AI features temporarily

If you need to turn off all AI features across DPMS without deleting your credentials — for example, during an audit period or while investigating an unexpected cost spike — navigate to AI Credentials and click Edit. Toggle the Activated switch to the off position (grey). Click Save. The credentials remain stored, but every AI button and auto-generation option across the platform is immediately disabled for all users in the current browser session and for anyone who loads any DPMS page after the save. Re-enable AI at any time by returning here and toggling Activated back on.

Connecting via Google Vertex AI

If your organisation routes Google AI usage through your own Google Cloud infrastructure (Vertex AI) rather than the public API, select Google from the AI Provider dropdown. A toggle labelled with the Vertex AI option will appear in the form. Enable that toggle: two additional fields — Project and Location — will appear beneath it. Enter your Google Cloud project ID and the compute region (for example, us-central1). Fill in any other required credential fields, select your default model, enable the Activated toggle, and click Save. If you later turn the Vertex AI toggle off and save again, the Project and Location values are excluded from the submission even if they still appear filled in — this is by design.

Checking the current AI configuration (read-only)

If you are a compliance officer or risk manager who needs to verify which AI model is active before relying on AI-generated content in a DPIA or vendor assessment, navigate to IT Settings › AI › Credentials. The summary card gives you everything you need at a glance: the provider name, the model name, the context window, the max output, and whether the integration is active or inactive. You do not need edit access for this — read access is sufficient, and the Edit button will show an access-denied page if you try to use it without the appropriate permission.

Field reference

AI Provider — The external company whose API DPMS will call. Choose from the available options in the dropdown. Required. Changing this field resets all other fields immediately.
API Key — The secret authentication credential issued by your AI provider. When editing an existing credential, the field initially shows a masked preview (for example, sk-***). Click the field only if you want to replace the key; if you leave it untouched, the stored key is re-submitted. Required for all providers.
Default Model — The specific AI model DPMS will use for every AI-assisted operation. Once selected, the context window and max output sizes for that model are displayed below the field. Required for all providers.
Reasoning Effort (OpenAI only, certain models) — Appears automatically when you select an OpenAI model that supports extended thinking. Disappears if you switch to a model that does not support this feature. Optional; the field is omitted from the submission if the selected model does not support it.
Vertex AI toggle (Google only) — Switches Google authentication from the standard public API to your organisation's Google Cloud infrastructure. When enabled, the Project and Location fields become required.
Project (Google Vertex AI only) — Your Google Cloud project ID. Required only when the Vertex AI toggle is on.
Location (Google Vertex AI only) — The Google Cloud compute region where your AI workloads run. Example: us-central1. Required only when the Vertex AI toggle is on.
Activated — Controls whether the saved credential is the live AI integration for the entire platform. Must be toggled on for AI features to be available. Defaults to off whenever you switch providers. Required in the sense that leaving it off means AI remains disabled even if all other fields are valid.

How this connects to the rest of DPMS

The AI Credentials screen is a prerequisite for virtually every AI-assisted feature in Priverion DPMS. No other screen sends you here automatically — you navigate here intentionally through the IT Settings menu — but the downstream effects of what you configure here are felt everywhere.

After you save an active credential, two things happen immediately across the platform: a flag is set that tells every module whether AI is available, and the context window size for your selected model is stored so that modules which send long documents to the AI can warn users if their content is too large for the model to process in one request. In practical terms, this means AI helper buttons on ROPA records, vendor questionnaires, asset descriptions, DPIA forms, and similar screens will become active or inactive based solely on what is saved here.

The platform stores only one AI credential configuration at a time. When you save new credentials, the previous configuration is replaced. If other users are actively using AI features at the moment you switch providers, they may briefly experience errors or see AI buttons become unavailable until the changeover completes. Plan provider switches during a low-activity window if possible.

After completing this screen, you may want to verify the AI integration is working by visiting any screen with an AI helper button — for example, opening a ROPA record and attempting to auto-generate a description. If AI features do not appear after saving an active credential, double-check that the Activated toggle was on when you saved.

Tips & common pitfalls

Heads up: Changing the AI Provider dropdown instantly clears all other fields with no warning and no undo. Always select your provider first, before typing your API key or choosing a model.

Tip: To update only the default model without touching the API key, open the edit form, change the model dropdown, and click Save without clicking the API key field. The masked key will be re-submitted as-is, so you won't need to re-enter it.

The Activated toggle resets to off every time you switch providers. After filling in all the fields for a new provider, remember to turn Activated back on before saving — otherwise your AI features will stay disabled even though valid credentials are stored.
Token limits are informational, not a guarantee. The Context Window and Max Output figures shown beneath the Default Model dropdown come from a reference table on the server. They may not reflect any usage caps or tier restrictions on your actual API subscription. Always verify your access level in your provider's developer console.
Only one credential configuration exists at a time. Saving new credentials replaces the previous ones entirely. There is no version history or rollback within DPMS.
Read-only access is sufficient for auditors. Compliance officers and auditors who only need to verify the active model and its capabilities can navigate to this screen with read permission alone. They will see all the relevant information on the summary card without being able to make changes.