Privacy and Security Controls and Methodology in the New AI Era (M02c)
AI has witnessed a significant surge in mainstream popularity, attracting substantial interest from venture capital and the media. This growth has been accompanied by a notable increase in security concerns. While literature addressing these concerns existed before ChatGPT, the subsequent focus on transformers and large language models (LLMs) has altered the landscape of their applicability, even for those who were already well-versed in these techniques.
In this talk, the speaker will delve into the present-day exploitation of AI-targeted advanced persistent threats (APTs) around the world. These threats encompass various areas, including credit risk fraud, insurance fraud, identity theft, and even illegal poaching. The talk will emphasize the intricate nature of these APTs, making them less susceptible to evasion compared to previous cyber trends. Furthermore, the talk will systematize cutting-edge methods for certifying AI models and safeguarding privacy. The speaker will argue that these methods merit serious consideration for regulatory purposes.