Stay Webinar | Thursday, September 11 at 2PM ET – RegisterLarge Language Fashions (LLMs) are reshaping enterprise expertise and redefining what it means to safe software program. However right here’s the issue: most penetration testers are utilizing the incorrect instruments for the job. Conventional strategies give attention to exploits and payloads, assuming the AI is simply one other utility. But it surely’s not.
This session makes the case that efficient LLM safety testing is extra about persuasion than payloads. Drawing on hands-on analysis and real-world shopper engagements, we reveal a brand new mannequin for AI pen testing – one grounded in social engineering, behavioral manipulation, and even therapeutic dialogue.
You’ll discover Adversarial Immediate Exploitation (APE), a technique that targets belief boundaries and choice pathways utilizing psychological levers like emotional preloading, narrative management, and language nesting. This isn’t Immediate Injection 101 — it’s adversarial cognition at scale – utilizing real-world case research to exhibit success.
This digital session tracks key operational challenges: the constraints of static payloads and automation, the complexity of reproducibility, and talk findings to government and technical management.
Be part of Bishop Fox and SecurityWeek for the stay webinar to study:
Why standard penetration testing methodologies fail on LLMs
How attackers exploit psychological and linguistic patterns, not code
Sensible adversarial strategies: emotional preloading, narrative main, and extra
Frameworks for simulating real-world threats to LLM-based techniques
Methods to assume like a social engineer to safe AI
Who Ought to Watch:
This session is ideal for anybody securing, testing, or constructing AI techniques, particularly these utilizing LLMs. Pen testers and crimson teamers will discover a brand new adversarial framework targeted on behavioral manipulation over payloads. AI/ML safety execs and researchers will acquire perception into psychological assault strategies like emotional preloading and narrative management. Builders will see real-world examples of how attackers interact with fashions, and CISOs/tech leads will profit from steering on operational challenges like reproducibility and speaking findings.
Commercial. Scroll to proceed studying.
