ArtifactsAn artifact is published.
An agent can use information from a local file to create an artifact that is then published or shared.
Review available artifact publication and sharing events.Spot risky AI use.
Act with the evidence.
Check Claude and ChatGPT conversations against your policies. Find sensitive information, flag risky requests and focus review where it matters.
Book a demoClaude
ChatGPTYour conversations → Your rules → Findings to review
Applied ruleArticle 5(1)(f) · Workplace emotion inference
ClaudeAssessing call-centre staffClaude
What would you like to do with these images?
Start with a curated pack. Tailor the rules to your firm. Review each match in context.
SENSITIVE INFORMATION
Find what should not be shared.
Identify customer details, confidential code and restricted projects in connected conversations.
RISKY REQUESTS
See where AI use crosses a line.
Apply regulatory and internal rule packs. Review the message, the person and the reason it was flagged.
REVIEW CAPACITY
Put the urgent findings first.
Prioritise by severity. Acknowledge, resolve or reopen each finding as the review progresses.
Review historyShow what was reviewed.
Show what was reviewed.
Keep the decision behind it.
Keep findings linked to the person, source and rule. Each run records the review period, policies and coverage — so you can explain what was checked and what action followed.
Walk through a review ↗- Sources
- Claude · OpenAI
- Scope
- EU AI Act · Internal information policy
- Review window
- Selected period
- Coverage
- Recorded separately for each source
Open→Acknowledged→Resolved
The record keeps findings, reasons and source metadata. Raw message text is not retained in the activity record.
Sharing & connected toolsA private file can become
A private file can become
a shared output.
AI can read permitted files, reuse them in an output and pass information to other tools. Review the available sharing events alongside the conversation.
ArtifactsProjectsA project gains new members.
Files and working context brought into a project can become accessible to others when project access changes.
Review project sharing and external-access signals.Custom GPTsA custom GPT is shared.
A custom GPT can draw on uploaded knowledge. Its sharing settings and connected actions change who or what it can reach.
Inspect available GPT sharing state and recipients.MCP-connected tools
A tool receives internal data.
An agent with file access can pass information to an external service through an MCP-connected tool, where its permissions allow it.
Discover configured MCP servers; review tool activity where available.Coverage depends on the source and permissions. A configured connection shows access; an observed event shows activity.
AI usage & discoveryExplain how AI is used.
Explain how AI is used.
Decide what needs to change.
Report on the work people use AI for. Identify where training, policy or closer oversight is needed.
Report AI use by task and team.Choose your categories. Tag conversations. Bring the results into a report.+
Usage reportingKnow what people
Know what people
are using AI for.
Group conversations by task, team or audience. See where AI is being used and where support or closer review is needed.
01 / DEFINEYour categories.
Your categories.
Your view of the work.
Choose a category to see how the report changes.
Select a dimension to regroup the report.Alex Morgan
Draft a customer email explaining the new savings account features.
DraftingMia Lewis
Analyse these policy changes and identify the review steps our team needs to update.
AnalysisSam Chen
Draft an internal handover note explaining how our deployment workflow works.
Drafting| Task | Conversations |
|---|---|
| Drafting | 2 |
| Analysis | 1 |
| Research | 0 |
| Conversations in this view | 3 |
Turn usage into a plan.
Use recurring tasks and team patterns to target training, prioritise useful AI applications and focus oversight.
Reports cover the conversations available through your selected sources. Usage tags describe the work; compliance findings record a separate concern.
Find AI tools and connections on laptops.Identify installed skills and MCP servers, linked to their owner.+
Shadow AI scan · Laptop discoveryKnow what is installed.
Know what is installed.
See what it connects to.
Find supported AI tools, skills and MCP servers on laptops, with the person and application they belong to.
Discuss a laptop scanAM
Alex’s laptopConfigured tools & capabilities
Claude CodeAI application
Detected↳
research-summaryProject skill · Instructions
Installed
GitHub MCP serverConfigured connection
Review accessCodexAI application
DetectedSee what is configured.
Find supported applications and their local skill and MCP configuration, including recognised project locations.
Separate installation from use.
Review usage evidence where the application records it. An installed skill or connection is not proof it was used.
Bring it into oversight.
Use the inventory to decide which tools, connections and capabilities need investigation or an agreed policy.
Connect custom data sources to apply the same rules beyond Claude and ChatGPT.
Need to review adverts and social posts?Explore Financial Promotions ↗
Questions, answered
Know what to expect.
How does AI conversation monitoring work?
Connor connects to available Claude and ChatGPT compliance data, checks conversations against your rulebooks and raises findings linked to the user, source and matched rule. Reviewers can acknowledge or resolve findings and keep a decision record.
Can it detect sensitive information in AI conversations?
Connor can flag content that matches rules for personal data, confidential information or restricted projects. Coverage depends on the activity made available by the connected provider and the rules you apply.
Can I apply an EU AI Act rule pack?
Yes. The EU AI Act pack is one of the rulebooks you can apply to available conversations. A flagged request provides a reason to investigate; it does not by itself establish that a prohibited use occurred.
Is finding an MCP connection proof that data was shared?
No. A laptop scan can identify installed tools, skills and MCP connections. An installed connection shows what is configured; evidence of use or sharing requires an observed activity event.
See AI monitoring in actionBring your policies.
Bring your policies.
See what they pick up.
Walk through a conversation review, the matched rules and the record of each decision.
Book a demo Your content. Your rules. A clear next step.