Cyber risk

AI-assisted forensics

Find what the evidence can support.

Bring scattered records into a reviewable account. We build an internal capability for approved read-only investigation. People review the findings and authorise response actions.

Follow an investigation question
Three red fragments with a continuous recessed seam. The fragments can be separated without concealing the missing material.
A connection is worth examining. A gap still matters.
Examine the fragmentsReturn to the first view

An abstract study, not an analysis result.

An illustrative disputed sign-in, not a customer case

“Was that sign-in mine?”

An employee disputes a sign-in. The record raises a question; it does not, by itself, establish who acted or what happened next.

  1. Records

    Bring approved identity, endpoint and cloud activity records together. Preserve references so reviewers can return to the source evidence.

  2. Question

    Examine the sign-in and subsequent activity within the agreed period. Compare competing explanations and show where records conflict or are missing.

  3. Reviewable account

    Prepare a timeline, explanation and evidence index. Accountable people check the findings, decide whether more evidence is needed and authorise any response.

Read-only investigation. Human authority. Isolation, remote execution, disruptive collection, remediation, wider scope and incident closure stay with authorised people.

Evidence sources and permissions: agree the boundary first

Your team defines the question, affected systems, time period, access boundary and approvals. Our engineers connect approved identity records, endpoint logs, cloud activity and existing security tools as the case requires. A connection does not guarantee that a source contains the answer.

The capability plans queries and hypotheses, correlates evidence and drafts findings within the agreed read-only scope. The model, credentials, case data, logs and outputs run under customer control. Connected systems retain their own access and operating requirements.

Delivery and control: a capability your team can operate

Agree the first investigation mission, sources, autonomy boundary and success criteria. Deploy the selected model and orchestration on customer-controlled infrastructure, integrate sources and encode access, provenance and audit controls. Prepare runbooks and train the operators and reviewers.

Everything we build for you, including its IP, is yours. You can change the underlying model, decide who operates the capability and run it without us.

Evidence and legal limits: no guaranteed finding or incident retainer

Missing records can leave events or data exposure unresolved. Finding no evidence of an action is not proof it did not happen. AI-generated findings need human review against the underlying records.

Regulatory or legal matters require specific chain-of-custody controls, expert review and approval gates agreed with counsel or the relevant stakeholder. These govern evidence collection, preservation and handling, without guaranteeing court admissibility, regulatory acceptance or compliance certification.

This is an internal investigation capability, not an incident-response retainer. Specialist firms may still be needed for novel malware, independent validation, expert testimony or counsel-directed matters. Findings can inform risk assumptions but do not themselves establish a financial loss estimate.

How evidence informs a protection decision