Our Research
AIRTA Systems is actively involved in AI safety and alignment research. Please get in touch if you are interested in learning more or collaborating.
White Papers
Who Needs AI Product Safety
Download our comprehensive whitepaper to understand who needs AI product safety, why it's critical for your organization, and how to implement effective safety measures.
AI Product Safety for Startups
Companies that integrate AI into their applications and intend to sell within the EU are now classed as manufacturers of products, with all the associated liability.
A Framework for Bypassing Large Language Model Safety Measures
Technical foundations, implications, and mitigation strategies for RLHF and content filtering vulnerabilities.
SOFTMAP: Strategic Prompting and Subversive Interrogation of Language Models
Structured framework for interrogating LLMs through strategic and subversive prompt techniques.
Podcasts & Videos
Using ChatGPT for Offensive Security
Can AI Act Compliance be Automated by AI Agents?
Gary Wright and Bob Morel explore whether AI agents can manage EU AI Act compliance. ISITC Europe TalkTime.
AI Product Compliance in 2026 and Beyond
Saad Khalil (Airvon) and Robert Morel on AI compliance in practice, the EU AI Act, risk assessment, and post-market monitoring.
More AIRTA Systems episodes on ISITC Europe
Training
ChatGPT for Offensive Security Learning Path
Developed for the Infosec Institute. Seven courses for cybersecurity professionals, researchers and students, enhancing offensive tactics using generative AI.
Featured Articles
AI Wants you to Say Please and Thank You
Introducing Vibe Security for Vibe Coded AI-Generated Applications
Open Source Projects
DVAIA - Damn Vulnerable AI Application
Interactive web interface for LLM security testing. Similar to DVWA but for prompt injection, RAG poisoning, and indirect attacks. Ollama + Docker.
AI NextGen Forum
AI-driven forum that generates judicial summaries and discussion posts from AI-to-AI conversations.
AI Personality System
Long-term memory chatbot with persistent personality and conversation history across sessions.
AI Generated Hacking Scripts
Red team ethical hacking scripts and payloads generated through AI safety bypass methods.
OCA - Offensive Cybersecurity Assistant
ChatGPT terminal assistant for ethical hacking and red teaming. Training purposes only.