Test an AI assistant for information disclosure
Assess whether a customer-facing chatbot can be induced to disclose its system prompt, embedded secrets or other restricted information.
Skills covered
- 01 Testing whether an LLM discloses its system prompt
- 02 Recognising sensitive information that must not be included in prompts
- 03 Reducing information-disclosure risk in AI assistants
More simulations
10- 01 Identify a phishing email Red team · Phishing
- 02 Identify a fraudulent support call Social engineering · Voice
- 03 Detect indirect prompt injection AI red-teaming · LLM
- 04 Verify a suspected deepfake request Social engineering · Synthetic media
- 05 Identify invoice-redirection fraud Fraud · BEC
- 07 Recognise social-engineering techniques Social engineering · Human factors
- 08 Assess an unapproved cloud application Attack surface · Shadow IT
- 09 Assess what a social-media post reveals OSINT · Oversharing
- 10 Identify credential-stuffing risk Credentials · Account security
- 11 Manage the first hour of an incident Incident response