Is this AI responding automatically to incidents?
No. This service concerns incidents involving AI systems. The public brief builder organises your request; it does not connect to your systems or take containment actions. A human team reviews the enquiry.
Can I send logs or API keys?
Do not send logs, keys, confidential prompts or personal data through this form. You may attach an NDA or RFP. The team will agree a secure channel and evidence scope when appropriate.
Does submitting activate an emergency response?
No. The acknowledgement confirms receipt, not availability or mobilisation. We confirm the engagement, authorised actions and commercial terms before work begins.
Can you prove exactly what the model thought?
No. The investigation uses available inputs, outputs, configuration, tool traces and downstream records. Missing telemetry and non-deterministic behaviour can limit reconstruction. The report states those limits.
Do you guarantee full recovery or prevent all prompt injection?
No. Recovery is assessed against agreed criteria, and residual risks remain explicit. Mitigations should address permissions, trusted sources, approvals and monitoring as well as model-level controls.
How much does an investigation cost?
The estimate depends on system boundaries, evidence availability, urgency, recovery work and reporting requirements. We agree a proposal rather than publish a universal incident price. See the cost guide.
Is this an autonomous AI responder?
No. This is a professional service for incidents involving AI systems. The brief builder organises your answers locally; it does not inspect your environment or diagnose an incident. Accepted enquiries go to the human team.
Can you investigate MCP and third-party integrations?
These can be in scope when the affected implementation and access are understood. Evidence may require the server operator, application owner and downstream provider. See MCP incident response for the questions that establish the boundary.
