HackTalk with Zainab Majid
def/acc hackathon - HackTalk
build societal defenses against ai threats alongside 1000+ builders globally apart of the Defensive Acceleration Hackathon.
the talk
When AGI arrives, model weights become the most valuable intellectual property on Earth. Algorithmic secrets become weapons of civilizational consequence.
Traditional cybersecurity won't cut it. We need AI-native defenses built from first principles.
Zainab Ali Majid from Asymmetric Security will challenge you to rethink everything you know about defending critical systems. In a world of abundant intelligence, the rules change completely.
She'll cover:
First principles rethinking: What does cybersecurity even mean when both attackers and defenders have superhuman AI capabilities?
Model weight security: How do you protect the crown jewels when a single breach could hand nation-states or bad actors the keys to AGI?
Algorithmic secrets: The techniques, training methods, and architectural insights that matter as much as the weights themselves
Defensive cyber AI capabilities: What we need to build NOW to stand a chance of defending what's coming
the speaker
Zainab Ali Majid works at Asymmetric Security, focusing on AI-era cybersecurity challenges. She's thinking about the hard problems that emerge when intelligence becomes abundant and the attack surface expands exponentially - problems that current security paradigms aren't equipped to handle.
why this matters
You can't defend AGI systems with tools designed for the pre-AI era.
The stakes are existential: a breach of AGI model weights could accelerate dangerous capabilities proliferation by years. A compromised training pipeline could introduce undetectable backdoors. Traditional perimeter defenses crumble when attackers have AI agents probing for vulnerabilities at machine speed.
Zainab will show you what AI-native security actually looks like and give you the mental frameworks to build it.
hosted by bluedot impact and apart research. sponsored by Halcyon Futures