Prompt injection is an attack where text a model reads, such as a web page, email or document, carries instructions that override what the application intended. It is the main security risk for any model that can call tools.
An AI agent is software that wraps a language model in a loop with tools and memory so it can work toward a goal. It chooses an action, reads the result and continues until the task is done.
The OWASP Top 10 is an awareness document from the OWASP Foundation listing the most critical categories of web application security risk. It is revised periodically using community data.
AI security is protecting AI systems, their models, data and connected tools from attack and misuse. It also covers the new risks that AI adoption brings to the rest of the business.
Data poisoning is an attack that corrupts the data used to train, fine-tune or ground an AI model. The model then learns hidden flaws, biases or backdoors chosen by the attacker.
Guardrails are checks placed around a language model that inspect what goes in and what comes out, blocking or correcting responses that break rules. They enforce policy in code rather than trusting the prompt.
A clear proven process, that delivers
You've seen our ideas - now go deeper.
Discover more insights, guides, and engineering practice.