Markstedter challenged the cybersecurity community to develop reverse-engineering tools for AI and to reconsider concepts of identity access management in a world of autonomous systems. She emphasized the need to understand the technology that is changing our systems and threat models to address emerging problems. The keynote also included a surprise announcement of an AI Cyber Challenge by Perri Adams, program manager for DARPA’s Information Innovation Office, aimed at driving AI innovation and creating a new generation of AI-based cybersecurity tools.
Key takeaways:
- Maria Markstedter, founder of Azeria Labs, opened Black Hat USA with a keynote on the promise and perils of AI, warning of an emerging army of autonomous AI bots.
- Markstedter compared today's AI to the first-generation iPhone, stating that it's insecure and dominated by companies like OpenAI. She also highlighted the rise of multimodal AI, which pulls data from various sources, increasing the risk of data corruption.
- She envisioned the next big AI push to be autonomous AI agents that can process multimodal data inputs to generate consequential outcomes, and challenged the cybersecurity community to develop decompiling and reverse-engineering tools for AI.
- Markstedter emphasized the need for a reexamination of access management and data security in light of the increasing power and capabilities of AI systems, and the potential reality of autonomous AI agents within enterprises.