Harriet Farlow
CEO
Mileva Security Labs
Harriet Farlow is the Founder and CEO of Mileva Security Labs, specialising in AI and cybersecurity. A recognised leader and PhD researcher in machine learning security, she’s dedicated to making AI systems safer through research, tools, and advocacy
Speaker sessions
Cyber security is a subset of AI security: our AI security incident research on why this is the case
With AI becoming more deeply embedded in critical infrastructure, cyber operations, and geopolitical conflicts, we set out to answer a simple question - how likely are AI security incidents? After a year of investigation - in partnership with the Australian National University, University of New South Wales and generously funded by Foresight - by combing through AI incident databases, enriching the 91 AI security incidents recorded, and applying different statistical modelling techniques we don’t just come away with a number of insights into the AI security threat landscape today but we also have a hot take: cyber security will soon be a subset of AI security and not the other way round. In this talk, we defend this stance and present for the first time our research findings from this project. We apply a national security lens to this work to discuss key trends, such as how threat actors are operationalizing adversarial ML techniques in novel ways - for example, evading malware classifiers with poisoned data, embedding malicious payloads in open-source AI models, and more. We will also explore why current cybersecurity frameworks are insufficient for AI threats, how AI security requires new risk models, and the impending national security impacts of AI security. We argue that AI security—not AI safety—should be the primary concern for security professionals.
- 15:20
- Wed
- 03 Dec
Stage:
Briefings 2
Sessions Type:
Presentation