PowerPoint presentation on HOW TO MAKE AI COMPLY as presented at BlueteamCon2023.

How are you adopting AI/ML into your enterprise? How hard can building an AI really be? What should you consider when introducing an AI model? Are you prepared for the consequences? How to do set guard rails so that your developers don't break you into jail?

This talk is a Treatise on AI/ML challenges for governance and strategic guidance for securing AI/ML scenarios within an enterprise for product research and development. Developed from years of experience learning how to approach AI/ML projects in Microsoft Research, I will attempt to shed light on some of the current thinking and best practices for AI models for internal and public use.

You can expect to learn about the explosion of technologies becoming available to industry and the challenges facing platform builders and enterprises now that these features are present.

We'll take a journey through the current regulatory landscape and share some considerations that Security governance programs should think about for their own AI/ML compliance.

Learn about how security fundamentals have not really changed but must be enhanced to deal with these new realities, including new perspectives on protecting the supply chain, approaching AI scenarios in threat modeling, building controls for resilience and the new demands of logging and auditing in real time for intentional behavior vs malicious behavior.

Then we wrap up with how to approach educating compliance organizations, your leaders, and your developers to be prepared to understand AI/ML risks and adopt AI with due diligence.