Hacked by Design? AI’s Overlooked Backdoors and What They Mean for Security

In an era of wars, political instability, and digital proxy battles, cybersecurity is no longer a technical afterthought but a frontline of geopolitical conflict. Governments, corporates, and startups alike are navigating a world where cybercriminals and hostile states exploit instability with unprecedented speed.

While much of today’s debate centers on AI’s ethics, bias, and performance, an equally urgent issue is often overlooked: the native vulnerabilities of AI applications themselves. Every AI system deployed creates new attack surfaces: they can be poisoned, injected, manipulated, or exfiltrated. Unlike traditional software, these vulnerabilities are harder to predict, patch, and audit, and when scaled inside critical systems, they multiply business risk.

Meanwhile, malicious actors face none of the regulatory or ethical constraints of legitimate players. They are free to leverage generative AI and quantum tools to automate and repeat attacks endlessly, placing defenders at a constant disadvantage. In this context, the track asks: Are developers building AI securely by design? What must companies keep in mind when deploying AI in-house? And how can we prepare for the convergence of AI-native vulnerabilities, quantum threats, and geopolitical instability?

 
 16:00 – 16:20 | Fireside Chat

Speakers:

• Katie Paxton-Fear, Staff Security Advocate, Semgrep
• Moderator:Louise Gookey, Stage Host, 4YFN

16:20 – 17:00 | Panel Discussion

Speakers:

• Jessica Souder, AI Security Lead, Palo Alto Networks
• Chema Alonso, VP, Head of International Development, Cloudflare
• Steven Latré, Chief AI Officer, Openchip
• Moderator:Andrea Garcia Beltran, Founder & Chair, CyberSpecs