cyberpolycrisis.com

LLMjacking: Customer service report

Expanded web report · Related PDF snapshot: 22 October 2026 · PDF page 11

Research reviewed 22 October 2026 · Evidence and editorial planning guidance are distinguished below.

Download PDF

Give accurate answers and preserve customer commitments

Distinguish a busy or unavailable AI feature from evidence of customer-account compromise. Give customers an alternative route when automated assistance is paused. Reports of unexpected prompts or activity should be sent to the service owner, not tested by frontline staff against production.

Your immediate priorities

  • Use the approved message for capacity restrictions and manual support.
  • Capture minimal case metadata for suspicious behavior and escalate it.

How to work through the decision

Use one approved, dated account of what is known, what is still being checked and when the next update is expected. Avoid diagnosing compromise from a customer’s description alone. Record the minimum information needed to route a case; passwords, payment card details and recovery codes do not belong in ordinary support chat. Track unresolved orders and promised follow-ups separately from general incident questions.

Evidence to ask for

A support fallback and an incident trend summary without raw sensitive prompts.

Can a customer still reach a person when the AI service has been deliberately restricted?

Work with the next person

Send the business manager a concise list of customer effects, unresolved commitments and recurring questions. Escalate new security evidence through the incident channel, preserving the original report without circulating sensitive material.

Continue with the business manager report for this topic.

A scenario to rehearse

Illustrative exercise, not an observed customer incident: An internal demonstration endpoint is shared publicly and begins generating provider charges overnight. The exercise measures whether usage alerts lead to a verified shutdown before the next billing review.

What a useful outcome looks like

A support fallback and an incident trend summary without raw sensitive prompts. Keep the decision dated and return to it if new evidence changes the scope.

Sources and further reading

What does this mean for your business?

Use the Cyber Health Review to turn the research into priorities for your own operating model.

Contact help@cyberpolycrisis.com