

Beyond Private Keys: Security in the AI Agent Era
About Event
As AI becomes increasingly embedded in financial infrastructure, the security boundary of digital asset exchanges is changing.
Recent asset theft incidents have shown that attackers do not always need to compromise private keys or break through traditional security defenses. Instead, trusted identities, internal tools, automated workflows, and AI systems themselves can become part of the attack path.
At the same time, AI is accelerating vulnerability discovery and exploitation, while enterprise AI Agents, MCP tools, and Skills are creating entirely new attack surfaces.
Hosted by Amazon Web Services (AWS) × LTP, this technical session will walk through the full attack chain:
How assets are stolen → How attackers use AI to find entry points → How enterprise AI Agents can be exploited → How to secure and govern Agents, MCP, and Skills
The session will feature four technical deep dives and two live demonstrations, combining real-world cases with practical approaches to AI security and governance.
Agenda
2:00 PM – 2:05 PM
Opening Remarks
2:05 PM – 2:50 PM
① Incident Review | The Private Key Wasn’t Compromised — So How Were the Assets Still Stolen?
2:50 PM – 3:30 PM
② AI on the Offensive | How Far Can AI Take Vulnerability Discovery & Exploitation?
3:30 PM – 3:45 PM
Tea Break
3:45 PM – 4:35 PM
③ Exploiting AI Agents | What Changes When Enterprises Give Agents Real Permissions?
4:35 PM – 5:20 PM
④ Governing AI Agents | Using Open-Source Agents, MCP & Skills Securely
5:20 PM – 5:30 PM
Closing & Open Discussion
Speakers
Xueyao Bai
FSI DNB Solutions Architect, Amazon Web Services
Yulong Zhang
FSI DNB Solutions Architect, Amazon Web Services
Zhang Ping
GenAI Specialist Solutions Architect, Amazon Web Services
Note
The security incidents discussed during this session are based on publicly available information.
Certain AWS generative AI services referenced in this session are currently available in AWS Regions outside Mainland China. Service availability is subject to the relevant AWS regional websites.