Audit glossary

Jailbreaking

Tricking a generative AI system into ignoring its safety rules.

AAIA

Users craft prompts, role-plays or encodings that bypass the system’s restrictions to produce prohibited content or actions. It is related to prompt injection, which hides instructions in content the system processes.

On the exam

Defences are layered: model safety training, guardrails, output filtering, monitoring and red teaming.

Related terms

  • Prompt injectionText that tricks a language model into ignoring its instructions.
  • GuardrailsControls that constrain what an AI system accepts and produces.
  • AI red teamingDeliberately attacking an AI system to find weaknesses before others do.

AAIA Test yourself on this with timed AAIA practice questions in the AAIA Exam Simulation Hub, or read the free AAIA exam guide.

Test yourself with today’s question of the day.