Definition

Prompt injection

An attack that hides malicious instructions in input, like a web page or email, to trick an LLM into ignoring its rules or taking unintended actions.

How founders on the podcast use it

Feeding hidden instructions to an AI system through its inputs so it does something its operator never intended.

Insure Your AI Agent: This PhD Built The Policy

A security exploit where untrusted input manipulates an agent into ignoring its instructions. A key reason gateways matter — a prompt injection can bypass a policy but not a credit card limit.

Your Next Co-Founder Should Be AI

A cybersecurity vulnerability in AI systems where malicious instructions are embedded in input data to manipulate the model's behavior. One of the most important open security problems in enterprise AI deployment.

Medical Tourism Is a $100B Industry

Hear it on the podcast3 episodes

Insure Your AI Agent: This PhD Built The Policy Your Next Co-Founder Should Be AI Medical Tourism Is a $100B Industry

Learn the language of AI-native founders

Every week: the new terms, tools, and workflows founders are using, straight from the podcast.

Related terms in AI Fundamentals

Precision NudgingLirio's trademarked approach: behavioral interventions personalized to an individual's specific barriers, delivered with optimized content, channel, timing, and frequency.Predictive SurfacingThe AI capability to surface relevant information, context, or the next action before the user knows to ask for it. The difference between a reactive tool (responds to queries) and a proactive system (anticipates needs). The hallmark of genuinely AI native product design.Predictive SystemsTools that anticipate user needs based on data patterns.ProbabilisticProduces outputs based on likelihood, which can vary and can be wrong. This is how LLMs work.RAG (Retrieval-Augmented Generation)A technique where a model first looks up relevant information from documents or databases, then uses it to write a more accurate, up-to-date answer.RationalistsAn online community focused on reasoning and AI risk.Reasoning modelA model built to work through hard problems step by step before answering, trading extra time for better accuracy.Red teamingAdversarial testing that attacks your own systems to find vulnerabilities before someone else does.

Browse all 1,456 terms