Solutions · Security
Help teams move forward. Know what needs attention.
AI adoption is moving fast. Find the tools in browsers, on managed devices, and in connected records. Understand the business use, work with the owner, and choose a response that fits.
- Managed browserschat.deepseek.com · 10:24Observed
- Identity + provider recordsApproved-provider accessCompare
- Managed device monitorCursor · Codex CLI · GitHub MCPInstalled / configured
- Security sourcesapi.deepseek.com · SSEObserved
- Spend + uploadsAI subscription · expense recordPaid
DeepSeek
Browser + security evidenceBusiness use confirmed. Decision recorded.
In this example, the owner confirms public research. IT records a scoped exception for that use.
Managed device monitor
Find the tools installed beyond the browser.
AI use can start in a desktop app, a command-line tool, or an MCP configuration. Add that device evidence to the same discovery record your team already reviews.
Review setup and coverage- Installed tools
- Surface supported AI apps and developer tools, such as Cursor or Codex CLI, on monitored devices.
- MCP evidence
- Identify supported MCP installations and configurations for review. Keep the originating device and evidence source attached.
- What it means
- Installed or configured is a finding to investigate. It does not establish that the tool ran, which data it accessed, or that Tallin controls its execution.
- Content boundary
- The device monitor does not collect prompts or responses. Optional browser prompt and file monitoring is a separate setting controlled by each client’s workspace administrator.
The response workflow
Turn an unknown service into an owned review.
- Discover
- Review supported service visits from managed Chrome and Edge profiles, plus configured identity, CASB, SSE, SIEM, expense, and provider evidence.
- Establish context
- Compare approved access and available ownership records. A website visit alone does not prove which provider account was active or what content was entered.
- Choose the response
- Assign review, update sanctioned-use status, use a supported provider access action, or route a selected workload through Gateway controls.
- Retain the decision
- Keep the source, owner, disposition, and follow-up with the finding. Bring material gaps into the next project or policy review.
Review versus blocking
Be specific about the control.
Seeing a record, reviewing content, and blocking a request are different actions. Tallin shows the relevant source, permission, and request path.
- Discover
- Website visibility only is the default. It records supported AI website visits and managed browser context without inspecting prompts or files.
- Optional browser monitoring
- Each client’s workspace administrator can enable prompt and file monitoring on supported AI chat pages. Review redacted excerpts, file details, and potential sensitive-information findings in that workspace. Monitoring does not block prompts or uploads.
- Supervise
- Review retained content from supported enterprise compliance APIs with authorized, reason-gated reviewer access. This is not in-flight content filtering.
- Gateway
- Enforce approved models, spend limits, rate limits, and actor access on routed traffic. Content-level PII blocking is on the roadmap.
- Outside the path
- Connected provider APIs can observe direct calls, but Gateway cannot block those calls or tool and MCP execution side effects outside Tallin.
Data handling
Choose the deployment and capture mode.
- Hosted gateway
- Metadata-only capture stores no prompt or response content, but Tallin still processes routed content in memory. Full-content storage is an explicit capture option.
- Customer VPC
- The gateway processes traffic in your environment and sends requests to your configured AI provider. Tallin receives agreed content-free metadata, not prompts or responses.
- Access and archive
- Supervision archive placement and reviewer permissions are configured separately from gateway capture. Confirm the requirements for each service during setup.
Start with one source and one clear response.
Find a signal, confirm its owner and scope, then choose the review or control that matches the risk.