When something goes wrong in SecureAI, the error you see usually points to one of a few root causes: a model communication failure, an expired session, a workspace access problem, or an integration misconfiguration. This article covers the most common errors, what causes them, and how to fix them.
Model Not Responding
These errors appear when SecureAI cannot get a response from the AI model provider.
"Model unavailable" or no response after sending a message
The upstream model provider (OpenAI, Anthropic, Google, or a local model) is not reachable.
Possible causes:
- The provider is experiencing an outage.
- Your organization's API key for that provider has expired or been revoked.
- A network firewall is blocking outbound requests to the provider's API endpoint.
What to do:
- Try a different model from the model selector dropdown. If one provider is down, another may still work.
- Check the provider's status page (e.g., status.openai.com, status.anthropic.com) for known outages.
- If no models work, ask your administrator to verify that API keys are valid in Admin Panel > Settings > Connections.
- If you are on a corporate network, confirm with your IT team that outbound HTTPS traffic to AI provider endpoints is allowed.
"Context length exceeded" or message fails silently
Your message plus the conversation history exceeds the model's token limit.
Possible causes:
- A long conversation has accumulated too many tokens.
- An assistant with a large system prompt and attached knowledge bases is consuming most of the context window.
- You pasted a very large document into the chat.
What to do:
- Start a new conversation to reset the context.
- Shorten your message or break it into smaller parts.
- If using an assistant with knowledge bases, the retrieved context adds to the token count. Try asking your question without the assistant to see if the issue is context size.
- For large documents, upload them to a knowledge base instead of pasting them into the chat. Knowledge base retrieval pulls only relevant chunks rather than the entire document.
"Rate limit exceeded"
Your organization has hit the usage limit for the billing period, or the model provider is throttling requests.
Possible causes:
- Your team's token allocation for the current billing period is exhausted.
- Too many concurrent requests are hitting the provider's rate limit.
What to do:
- Ask your administrator to check usage in Admin Panel > Dashboard or Admin Panel > Users.
- Wait a few minutes and try again -- provider rate limits are usually temporary.
- Switch to a different model that may have separate rate limits.
For more detail on model issues, see Why is the AI model not responding?.
Session Expiry
These errors occur when your login session has timed out or been invalidated.
"Session expired" or unexpected redirect to the login page
SecureAI sessions have a configurable timeout. When your session expires, you are redirected to the login page.
Possible causes:
- You were inactive for longer than the session timeout period.
- An administrator changed authentication settings, which invalidated active sessions.
- Your browser cleared cookies or you are using a privacy mode that does not persist sessions.
What to do:
- Log in again. Your conversations and data are preserved -- only the session token expired.
- If this happens frequently, ask your administrator to increase the session timeout in Admin Panel > Settings > General.
- Check that your browser is not configured to clear cookies on close for the SecureAI domain.
- If you are using a private/incognito window, sessions will not persist after closing the window.
"Unauthorized" or "401" error mid-conversation
Your session token became invalid while you were actively using SecureAI.
Possible causes:
- The server was restarted, which can invalidate session tokens.
- Your account permissions were changed by an administrator.
- A token refresh failed due to a network interruption.
What to do:
- Refresh the page. If the session can be restored, SecureAI will do so automatically.
- If refreshing does not work, log out and log back in.
- Your in-progress message may not have been saved. Check your conversation history after logging back in -- if the last message is missing, resend it.
Workspace Access Errors
These errors relate to permissions and access control within SecureAI workspaces.
"You do not have access to this workspace"
You are trying to access a workspace that your account is not a member of.
Possible causes:
- You were removed from the workspace by an administrator.
- The workspace was deleted or renamed.
- You are logged into the wrong account (if you have multiple accounts).
What to do:
- Verify you are logged into the correct account by clicking your profile icon.
- Ask your administrator to confirm your workspace membership in Admin Panel > Users.
- If the workspace was recently restructured, you may need a new invitation.
"Permission denied" when accessing a feature
You can see the feature in the interface but cannot use it.
Possible causes:
- Your user role does not include the required permission. SecureAI uses role-based access control with roles such as user, moderator, and admin.
- The feature was recently restricted by an administrator.
What to do:
- Check your current role by clicking your profile icon and looking at your account settings.
- Ask your administrator to grant the necessary permissions if your role requires the feature.
- Common features that require elevated permissions:
- Creating shared assistants (may require moderator or admin role)
- Accessing the admin panel (admin role only)
- Managing knowledge bases (may be restricted by workspace settings)
- Viewing other users' conversations (admin role only)
"Knowledge base not found" or missing search results
A knowledge base you previously used is no longer accessible.
Possible causes:
- The knowledge base was deleted by its owner or an administrator.
- The knowledge base's sharing settings were changed.
- The knowledge base is still indexing after a recent upload.
What to do:
- Navigate to Workspace > Knowledge to see your available knowledge bases.
- If the knowledge base was shared with you by a colleague, ask them whether it still exists.
- If you recently uploaded documents to a knowledge base, wait a few minutes for indexing to complete. Large uploads can take several minutes to process.
Integration Failures
These errors occur when SecureAI's connections to external services break.
"Integration connection failed" or sync errors
An external service integration (HubSpot, Google Drive, Slack, Microsoft 365) is not working.
Possible causes:
- The integration's OAuth token has expired and needs to be reauthorized.
- The external service changed its API or revoked access.
- Network connectivity to the external service is blocked.
What to do:
- Go to Settings > Integrations (or ask your admin to check Admin Panel > Integrations).
- Look for a warning icon or "disconnected" status on the affected integration.
- Click Reconnect or Reauthorize to refresh the OAuth token.
- If reconnection fails, the external service may be experiencing downtime. Check the service's status page.
- For corporate networks, verify that outbound access to the integration's API endpoints is not blocked by a firewall.
"Tool execution failed" in an assistant response
An assistant tried to call an external tool or API and the call failed.
Possible causes:
- The tool's endpoint URL is incorrect or unreachable.
- The tool requires authentication credentials that have expired.
- The tool returned an unexpected response format.
- The external service the tool connects to is down.
What to do:
- Try the same question without the assistant (direct model chat) to confirm the model itself is working.
- Report the error to your administrator -- tool configurations are managed in the admin panel.
- If you are an admin, check the tool configuration in Admin Panel > Tools. Verify the endpoint URL, authentication headers, and expected response format. See the tool integration guide for setup details.
"File upload failed" or "Import error"
An integration tried to pull in a file from an external service but failed.
Possible causes:
- The file exceeds the maximum upload size (check your deployment's configured limit).
- The file format is not supported.
- The integration's access permissions do not include read access to the specific file or folder.
What to do:
- Check the file size against your organization's upload limit. The default maximum is displayed on the upload interface.
- Verify the file format is supported. See Supported file formats for the full list.
- For integration-sourced files, ensure the integration has read access to the file's location in the external service.
General Troubleshooting Steps
When you encounter an error not listed above, try these steps in order:
- Refresh the page -- a hard refresh (Ctrl+Shift+R or Cmd+Shift+R) clears cached scripts that may be causing the issue.
- Try a different browser -- this rules out browser-specific issues, extensions, or cached state.
- Check your network -- confirm you can reach other websites. If you are on a VPN, try disconnecting to rule out VPN-related routing issues.
- Clear browser data -- clear cookies and cache for the SecureAI domain specifically, then log in again.
- Check for announcements -- your administrator may have posted maintenance notices or known issue alerts.
- Contact support -- if the problem persists, provide:
- The exact error message (text or screenshot)
- What you were doing when the error occurred
- The model and assistant you were using (if applicable)
- Your browser and operating system
- Whether other users are affected
Related Articles
- Why is the AI model not responding? -- detailed model troubleshooting
- What AI models are supported? -- available models and providers
- How is SecureAI billed? -- plan limits and usage tracking
- Can assistants call APIs and use tools? -- troubleshooting tool configurations
- Can SecureAI connect to Slack, Microsoft 365, or Google Drive? -- integration setup