An AI attack pattern which starts with innocent prompts and gradually escalates over multiple conversation turns to bypass safety guardrails and extract harmful content. Each individual turn appears benign, evading single-prompt detection.
Multi-turn escalation
7th May 2026
Centralized execution, reporting and insights in one place.
Explore MoT
Learn how to structure and organize test management data as an AI-ready foundation that gives AI systems richer context for testing insights and workflows.
Boost your career in software testing with the MoT Software Testing Essentials Certificate. Learn essential skills, from basic testing techniques to advanced risk analysis, crafted by industry experts.
Debrief the week in Quality via a community radio show hosted by Simon Tomes and members of the community
