Known Issues

Overview

We're committed to transparency about ISMS Copilot's current limitations and the issues we're actively working to resolve. This page documents known issues reported by users, explains what we're doing to fix them, and provides workarounds where available.

Who This Is For

This article is for:

• Users experiencing technical issues with the platform

• Anyone wanting to understand current platform limitations

• Teams evaluating ISMS Copilot who need transparency about known issues

• Users looking for workarounds while we work on permanent fixes

Current Service Status

For real-time service health and incident updates, check our Status Page. This page shows current uptime, ongoing incidents, and scheduled maintenance.

AI Provider Failover Notifications

ISMS Copilot uses Anthropic (Claude) as the default AI provider. When Anthropic experiences outages or connectivity issues, the platform automatically switches to OpenAI to keep your compliance work uninterrupted.

What You'll See During Failover

When the platform switches providers, an orange banner appears below the header in your chat workspace:

"Our default AI provider is experiencing an outage. We have temporarily switched to another provider to keep the assistants available. Sorry for the inconvenience — we will switch back once the default provider is online."

The banner includes a link to the Status Page where you can monitor the current situation and view uptime history.

Does This Affect Data Security?

No. Failover does not change ISMS Copilot's core security guarantees:

• EU-hosted data: Your conversations, files, and workspace data remain in Frankfurt, Germany

• End-to-end encryption: All data in transit and at rest stays encrypted

• Zero training on your data: Neither provider uses your inputs for AI training

• GDPR compliance: Standard Contractual Clauses (SCCs) apply to both Anthropic and OpenAI

Both providers are US-based and process queries under the same contractual safeguards. If you require EU-only AI processing for strict data residency requirements, enable Advanced Data Protection Mode, which uses Mistral (EU-based) exclusively and disables automatic failover.

How to Check AI Provider Status

If you're experiencing disruptions or want to verify the current provider status:

1. Check for the orange banner in your chat workspace—it appears automatically during failover

2. Visit the Status Page: User Menu (avatar, top right) > Help Center > Status, or go directly to status.ismscopilot.com

3. Review your Data Protection settings: Settings > Data Protection tab shows your current mode (Standard or Advanced)

When Does the System Switch Back?

The platform continuously monitors Anthropic's health and automatically switches back once the service is stable. This typically happens within minutes to hours, depending on the outage duration. The orange banner disappears when the switch-back completes.

First-Request Errors During Failover

During the initial moments of a failover (when Anthropic first goes down), you may briefly see "AI service unavailable" errors before the backup provider activates. If this happens:

• Wait 30-60 seconds for the failover to complete

• Retry your request—it should work once OpenAI is active

• Check the orange banner to confirm failover is in progress

This brief interruption is normal as the circuit breaker detects the outage and activates the backup provider.

Active Issues

Chat Session Timeouts During Long Queries

Issue: Very complex queries or document generation requests may timeout if they exceed standard processing limits.

Status: Significantly improved in December 2025 with extended chat timeout to 10 minutes (soft) / 20 minutes (hard). Most long responses now complete successfully.

Who may still be affected:

• Users requesting extremely comprehensive policy documents (20+ pages)

• Users running simultaneous complex queries across multiple workspaces

• Users with slower network connections during high server load

Workaround (if you still experience timeouts):

1. Break large requests into sections: Request one policy section at a time instead of entire manuals

2. Retry if needed: The extended timeout makes retries much more likely to succeed

3. Use the resume feature: Type "continue" or "you didn't finish" and the AI will complete the response

AI Accuracy & Limitations

AI May Provide Incorrect Information

Issue: The AI assistant can occasionally provide inaccurate guidance or incorrect control references, even though it's trained on real-world ISO 27001 implementation knowledge.

Why this happens: Like all AI systems, ISMS Copilot is probabilistic and can make mistakes. This is particularly true for:

• Specific control numbers and exact standard text

• Edge cases not well-represented in training data

• Questions that require very recent regulatory updates

• Complex multi-framework compliance scenarios

What we're doing:

• Continuously refining the AI's system prompts to improve accuracy (recent update in October 2025 made responses more concise and reliable)

• Expanding our training knowledge base with more real-world implementation examples

• Building verification mechanisms to flag uncertain responses

Workaround:

• Cross-reference AI suggestions with official ISO standards purchased from ISO.org

• Use specific, detailed questions (e.g., "How do I implement ISO 27001:2022 control A.8.1?" instead of "Tell me about access control")

• Review AI-generated policies and procedures with a compliance expert

• Test recommendations in your organizational context before formal implementation

File Processing Issues

Supported File Formats

Current limitation: ISMS Copilot only supports specific file types for upload.

Supported formats:

• Documents: PDF, DOC, DOCX

• Spreadsheets: XLS, XLSX, CSV

• Text: TXT, JSON

Not supported: Images (JPG, PNG), presentations (PPT, PPTX), videos, audio files, or executables

Rate Limiting & Performance

Daily Message Limit for Free Users

Issue: Free plan users see a clear upgrade dialog when reaching the daily message limit.

Why this happens: The free tier includes a limited number of messages per 4-hour rolling window to manage server costs and ensure fair access for all users.

This is intended behavior, not a bug. However, we understand it can be frustrating when you're in the middle of compliance work.

Updated in December 2025: The upgrade experience is now clearer with a dedicated dialog explaining your options and benefits of upgrading.

Options:

• Upgrade to Essential or higher (starting at $12/month): Get 50 credits per session and priority processing

• Wait for the limit to reset: The counter resets after 4 hours from your first message

• Optimize your questions: Ask comprehensive questions with full context to get more value from each message

Slow AI Response Times

Issue: Some users experience "AI response is taking longer than expected" messages during peak usage times.

Why this happens:

• High server load during peak hours (typically business hours in EU/US time zones)

• Complex questions requiring extensive knowledge retrieval

• Large file uploads being processed simultaneously

Status: We improved typing animation performance in October 2025 to make the interface feel more responsive. Backend performance optimization is ongoing.

Workaround:

• Wait 1-2 minutes before refreshing—most responses complete within this time

• Try during off-peak hours if possible (early morning or evening in your timezone)

• Break complex questions into simpler, more focused queries

• Check the Status Page for current service health

Authentication & Account Issues

Session Timeouts

Issue: Users occasionally see "Authentication failed. Please log in again" and must re-authenticate.

Why this happens: For security, authentication sessions expire after a period of inactivity. This is particularly common on shared or public computers.

This is expected security behavior, but we're working to balance security with user convenience.

Workaround:

• Save your work frequently in long sessions

• Use the "Keep me signed in" option when logging in (if on a trusted device)

• Refresh the page if you see authentication errors rather than logging out completely

OAuth Sign-In Inconsistencies

Issue: Occasional "Authentication Error" when signing in with Google or Microsoft accounts.

Why this happens:

• Browser cookie/cache conflicts with OAuth providers

• Corporate security policies blocking third-party authentication

• Expired OAuth consent requiring re-authorization

Workaround:

• Clear browser cookies and cache, then try again

• Use incognito/private mode to isolate authentication issues

• Try the alternative OAuth provider (switch between Google and Microsoft)

• Use email/password authentication as a fallback

• Check with your IT department if corporate firewalls may be blocking OAuth

Payment & Subscription Issues

Payment Processing Errors

Issue: Users occasionally encounter payment errors during upgrade or subscription renewal.

Common error messages:

• "Payment session not found" - Checkout session expired before completion

• "Payment was not completed successfully" - Card declined or payment method issue

• "User identification error during payment" - Authentication expired during checkout

• "Failed to update your account. Please contact support." - Backend update issue

Status: We updated our Stripe integration in October 2025 for more reliable payment processing. Most users no longer experience these issues.

Workaround:

• Retry the payment from a fresh session (log out and back in)

• Verify your payment method is valid and has sufficient funds

• Complete payment in one session without navigating away

• Contact support if errors persist—we can manually verify your subscription status

• Check your email for payment confirmation even if the UI shows an error

Subscription Status Not Updating

Issue: Upgrade shows as successful but interface still displays "Free" plan.

Why this happens: There can be a brief delay (up to 5 minutes) between payment completion and account status synchronization.

Workaround:

• Wait 5 minutes, then refresh the page

• Sign out and sign back in to force a status refresh

• Check your email—if you received payment confirmation, your upgrade is successful

• Verify your subscription in the Stripe customer portal (User Menu > Manage Subscription)

• Contact support if the status doesn't update after 15 minutes

Browser & Compatibility Issues

Interface Performance on Older Browsers

Issue: Users on outdated browsers may experience slow interface performance or display issues.

Recommended browsers:

• Chrome 100+ (best performance)

• Firefox 100+

• Safari 15+

• Edge 100+

Known issues with older browsers:

• Typing animation may appear sluggish (partially fixed in October 2025 for modern browsers)

• File drag-and-drop may not work consistently

• Some UI elements may not display correctly

Workaround:

• Update to the latest version of your browser

• Switch to Chrome for the most reliable experience

• Use the paperclip button for file uploads if drag-and-drop doesn't work

• Disable browser extensions that may interfere with web apps

Mobile Experience Limitations

Current limitation: ISMS Copilot is optimized for desktop use. Mobile experience has known limitations.

Known mobile issues:

• File upload interface less intuitive on small screens

• Long AI responses difficult to read without scrolling

• Document generation previews may not display properly

• Workspace management harder on mobile

Recommendation: Use ISMS Copilot on desktop/laptop computers for the best experience, especially when working with documents or complex compliance queries.

Workspace & Organization Issues

Workspace Data Retention

Issue: Maping sure your data doesn't suddently disappear because of too strict data retention settings.

Why this happens: Workspaces are subject to your data retention settings. By default, conversations and files older than your configured retention period are automatically deleted.

This is intended behavior based on your privacy settings, but it can be surprising if you weren't aware of the automatic deletion.

How to prevent data loss:

1. Go to User Menu > Settings

2. Check your Data Retention Period setting

3. Adjust to a longer period if you need to keep workspace data (maximum: 7 years)

4. Or select Keep Forever to prevent automatic deletion

5. Export important conversations before they expire

Limited Workspace Visibility in Sidebar

Issue: Not all workspaces appear in the sidebar navigation.

Why this happens: The sidebar only shows your 3 most recently used workspaces to keep the interface clean.

This is intentional design, not a bug. All your workspaces still exist.

How to access all workspaces:

• Click "View all workspaces" at the bottom of the sidebar

• This opens the complete workspace list where you can search, manage, and select any workspace

Beta Status & Evolving Features

ISMS Copilot Is in Active Development

ISMS Copilot is continuously evolving based on user feedback and compliance framework updates. Some features are still being refined.

What this means for users:

• New features are added regularly (check the Product Changelog)

• UI elements may change or improve over time

• Some edge cases may not be fully handled yet

• We actively collect feedback to prioritize improvements

Recently Fixed Issues

These issues were reported by users and have been resolved in recent updates:

✅ Long-Form Output Formatting and Truncation (Fixed: March 2026)

Issue: Very long AI responses like audit reports, comprehensive policy documents, and multi-framework gap analyses would sometimes lose formatting partway through or end abruptly without completing.

Status: Fixed in March 2026 by raising the output token limit from 8K to 64K tokens. Long-form content now generates reliably with consistent formatting throughout.

What improved:

• Audit reports and detailed policy reviews complete without truncation

• Tables, lists, and structured content maintain proper formatting in long documents

• Responses can now be up to 8x longer than before

✅ Conversation Context Loss in Long Chats (Fixed: March 2026)

Issue: In very long conversations, the AI would lose track of context from early messages — files uploaded at the start, key decisions, or audit findings discussed earlier.

Status: Fixed in March 2026. Automatic conversation compaction now works across all modes (Think and Fast). The system automatically summarizes earlier messages while preserving key information when approaching 100,000 tokens, so conversations can continue indefinitely without losing context.

✅ Generic Error on Token Limit (Fixed: March 2026)

Issue: Users would see a generic "AI service ran into an error" message when conversations exceeded the model's context window, with no explanation about why.

Status: Largely resolved by automatic conversation compaction (March 2026), which triggers at 100,000 tokens and prevents conversations from hitting the context limit in the first place. Users should no longer encounter this error during normal usage.

✅ Document Generation Failures (Fixed: March 2026)

Issue: PDF and DOCX exports would occasionally fail or produce truncated output, especially for long documents.

Status: Fixed in March 2026. Document generation now uses a markdown-based pipeline instead of AI-generated HTML, and the AI model was upgraded to Claude Sonnet 4.6. Even 20+ page policies covering all 93 Annex A controls now generate without truncation.

✅ Excel Files Freezing the Service (Fixed: March 2026)

Issue: Uploading oversized Excel files with millions of empty rows could freeze all chat machines, blocking all users on the platform.

Status: Fixed in March 2026. Oversized Excel files are now automatically stripped and capped, and a pre-flight size check prevents multi-minute hangs on oversized requests before they reach the AI.

✅ False Subscription Downgrades (Fixed: March 2026)

Issue: Some users experienced unexpected subscription downgrades when Stripe webhook events arrived out of order during plan changes.

Status: Fixed in March 2026. The subscription sync logic now correctly handles webhook race conditions and no longer triggers false downgrades.

✅ Better Error Handling When AI Service Is Busy (Fixed: March 2026)

Issue: If the AI service was temporarily overloaded during a response, you would lose the entire partial response and see a red error message.

Status: Fixed in March 2026. You now keep any partial response already received. The error message is clearer ("AI service is temporarily busy") and uses a softer amber warning instead of a red error.

✅ AI Provider Outages Affecting Responses (Fixed: December 2025)

Issue: When the primary AI provider (Anthropic) experienced outages, all AI responses would fail, interrupting compliance work.

Status: Fixed in December 2025 with automatic failover from Anthropic to OpenAI. The platform now seamlessly switches to the backup provider during outages without user disruption. Note: failover applies only to default mode; Advanced Data Protection Mode (EU-only via Mistral) does not have failover.

✅ Chat Loading Spinner Persisting After Messages (Fixed: December 2025)

Issue: Loading spinner sometimes remained visible after AI responses completed, creating confusion about response status.

Status: Fixed in December 2025. Loading indicators now properly clear when messages complete. January 2025 further improved the loading experience with a full-screen spinner and "Loading messages..." text during app initialization.

✅ Message Concatenation Errors (Fixed: December 2025)

Issue: In rare cases, multiple messages would merge together or display incorrectly in chat history.

Status: Fixed in December 2025 with improved message handling logic.

✅ Extended Chat Timeout (Improved: December 2025)

Issue: Complex queries would timeout before completion, requiring retries.

Status: Chat timeout significantly extended in December 2025 (now 10-minute soft timeout, 20-minute hard timeout), allowing most complex responses to complete successfully.

✅ Sluggish Typing Animation (Fixed: October 2025)

Issue: Chat typing animation felt slow and unresponsive.

Status: Fixed in October 2025. Messages now appear smoothly and responsively.

✅ Document Generation Reliability (Improved: October 2025)

Issue: Document generation frequently failed or produced errors.

Status: Significantly improved in October 2025 with better error handling and stability. Some edge cases may still occur.

✅ Verbose AI Responses (Improved: October 2025)

Issue: AI responses were too long and not action-oriented.

Status: Enhanced system prompt now delivers more concise, practical guidance while maintaining accuracy.

Reporting New Issues

When to Report an Issue

Please report issues if:

• You encounter errors not listed on this page

• Workarounds provided here don't resolve your problem

• You experience repeated failures of the same feature

• You discover incorrect compliance information that could impact audits

• You find security vulnerabilities or data privacy concerns

How to Report

1. Click your user avatar (top right)

2. Select Report Issue

3. Describe the problem with:

   • Exact error message (screenshot if possible)

   • Steps to reproduce the issue

   • Browser and device information

   • Date and time when it occurred

   • Workspace or conversation affected (if applicable)

4. Submit the form

Expected result: Our support team typically responds within 4 hours during business hours.

What's Next

• Troubleshooting Common Issues - General troubleshooting steps

• Service Status Page - Real-time service health

• Product Changelog - Latest features and fixes

• Trust Center - Security and compliance documentation

Getting Help

For questions or issues not covered here:

• Check the Status Page: isms-copilot.instatus.com

• Report an Issue: User Menu > Report Issue

• Browse Help Center: User Menu > Help Center

• Contact Support: We typically respond within 4 hours during business hours