• Skip to main content
  • Skip to header right navigation
  • Skip to site footer
MEST Center

MEST Center

National Microelectronic Security Training Center

  • Home
  • About Us
    • Highlights
    • Advisory Board
  • Contributors
  • Trainings & Schedule
    • Join us on nanoHUB!
    • Webinars
    • Micro Certificates
    • Macro Certificates
    • Other
      • Modules
  • Contact Us

Securing AI-Designed Silicon: New Risks, Hard Questions, and What We Must Do at Every Layer of the Stack

April 22, 2026 by Limor Herb

Date/Time
Date(s) - 04/22/2026
12:00 PM - 1:00 PM
Add to Google Calendar or iCal/Outlook Calendar


Speaker 

Mohamed Kassem is CEO and Co-Founder of NativeChips, a company building an AI-powered platform that turns natural language into manufactured silicon.

Abstract

AI-driven chip design is an operational reality. Production EDA tools use ML for synthesis and timing closure, LLMs generate synthesizable RTL from natural language, and multi-agent systems produce complete SoCs autonomously. AI-designed chips can be secure — the tools, IP libraries, and verification methods exist to build trustworthy silicon. The challenge is not the capability of the AI, but what we don’t yet know about its blind spots. AI systems optimize for functional correctness, not adversarial resistance. They can generate an AES module that encrypts and decrypts perfectly but has a weak key schedule, or a boot FSM that works 99.999% of the time but has a one-clock-cycle race condition. These are not failures of the tools — they are gaps in what the AI was trained to care about. This presentation examines those blind spots at every layer of the stack — architecture, RTL, verification, IP supply chain, physical design, and firmware — identifying risks unique to AI-generated hardware: prompt injection, training data poisoning, hallucinated security logic, correlated verification gaps, and adversarial AI Trojans. We then explore how AI itself is also the most promising tool for finding these blind spots — through ML-based Trojan detection, AI-guided formal verification, LLM-powered vulnerability scanning, and pre-silicon side-channel analysis — and identify the questions the community must ask to close the gap between what AI can build and what AI can guarantee.

Biography

Mohamed Kassem is CEO and Co-Founder of NativeChips, a company building an AI-powered platform that turns natural language into manufactured silicon. NativeChips automates the entire chip design flow using multi-agent AI — from a plain English description of what you need to a tapeout-ready design — making custom chips accessible to companies that were never able to afford them or have the expertise to make them before. Mohamed is also the founder of ChipFoundry, a community initiative making chip fabrication accessible to startups, researchers, and independent designers at a fraction of traditional cost. Before NativeChips, as CTO of Efabless, he built the chipIgnite platform and the open-source silicon infrastructure behind hundreds of community-driven tapeouts — helping turn open-source chip design from a research concept into real chips in real packages shipped to real people.

 



Registration

Tickets

Registration Information

an email from MEST
an email or newsletter from nanoHUB
LinkedIn post
MEST instructor or staff
Co-worker/colleagues
Supervisor/manager
other

Strongly agree
Agree
Neither agree nor disagree
Disagree
Strongly disagree

Strongly agree
Agree
Neither agree nor disagree
Disagree
Strongly disagree

Strongly agree
Agree
Neither agree nor disagree
Disagree
Strongly disagree

Strongly agree
Agree
Neither agree nor disagree
Disagree
Strongly disagree

Booking Summary

1
x Standard Ticket
$0.00
Total Price
$0.00

Stay in touch!

Join our LISTSERV

Join us on nanoHUB

  • LinkedIn
  • Email MEST Center
  • Join us on nanoHUB!