Contents
As AI systems are increasingly embedded in enterprise products and workflows, technical documentation is no longer a passive information artifact. Product documentation, example, and instructional content are now actively consumed by AI systems and can directly influence AI behaviour. This session examines prompt injection not as a coding or hacking issue, but as a technical communication challenge.
Using short, real world examples from enterprise documentation, such as API samples, troubleshooting steps, and error messages, the session explains how language written for human readers can be unintentionally interpreted by AI systems as executable instructions, priority overrides, or contextual commands. Participants will learn how everyday writing choices can introduce risk without malicious intent.
The session then introduces prompt-aware writing patterns, including instruction separation, content containment, and explicit boundary signalling. The talk reframes technical communication as contributors to AI safety, sytsem resilience, and trust, with guidance to apply AI enabled practices in documentation environments
Takeaways
Learn to identify prompt -injection risks in documentation, and apply prompt-aware writing techniques that improve clarity for users while supporting safer, and more predictable AI behaviour
Prior knowledge
No prior knowledge of AI security, prompt engineering, or programming is required; familiarity with AI‑enabled products or documentation environments is helpful but not mandatory.