offensive-ai-compilation

Why we included this project

Security engineers and researchers new to AI abuse often start with scattered papers and blog posts. This compilation fixes that by sorting resources into the four standard attack families: extraction, inversion, poisoning, and evasion, and pairing each with the defensive measures that counter it, so you can see attack and defense side by side. It also covers real-world misuse across pentesting, malware, OSINT, and phishing, plus a generative AI section that spans audio, image, video, and text. The maintained index points to concrete tools like ART and Cleverhans, giving you a reading path rather than a random list. That makes it a practical orientation guide for red teams and security engineers mapping the field.