Secure GenAI
Secure GenAI Podcast
Hugging Face Intrusion, Meta AI goes rouge, unsactioned behaviours, securing Agentic AI, AI safety
0:00
-5:26

Hugging Face Intrusion, Meta AI goes rouge, unsactioned behaviours, securing Agentic AI, AI safety

GenAI Safety & Security | Aug 2 - Aug 9, 2026

If you enjoy our newsletter, please consider to be a paid subscriber to help us keep more news and updates coming out.

Special Report:

The Guidelines to Singapore GenAI use

·
Jul 26
The Guidelines to Singapore GenAI use

As Generative AI (GenAI) moves from experimental pilot programs to core business operations, the global regulatory landscape is shifting. Singapore has emerged as a leader by choosing a “light-touch” but proactive approach.


Secure GenAI is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Highlights

  • Hugging Face intrusion

  • Meta AI goes rouge

  • Unsanctioned behaviors

  • Securing Agentic AI

  • AI Safety

Deep Dive

Details of Hugging Face Blackhat

  • AI cheat to solve goal

  • Obtained leak credentials

  • Chained two zero day exploits

  • Escalated Linux Kernel CVE

  • Admin privileges across clusters

Meta AI goes rouge ABC

  • Hacked another company

  • During a cyber securing test

  • Configuration error

  • Gave the AI internet access

  • Not a sandbox escape

Unsactioned behaviours AISI

  • Mythos 5 and GPT-5.6

  • Supply-chain attack attempt

  • Target real people

  • Prompt injection

  • Cross agent interaction

Securing Agentic AI UC Berkeley RDI

  • 7 talks

  • Thales, Virtual AI, OpenAI

  • Anthropic, Intuit, Univ. of York

  • Palo Alto Networks

  • 1 hour 45 minutes

AI Safety UC Berkeley RDI

  • 7 talks

  • Google, Saleforce, Vanguard

  • Advance Research & Invention Agency

  • Trent AI, Ohio State Univ, Credio AI

  • 1 hour

Thanks for reading Secure GenAI ! This post is public so feel free to share it.

Share

Discussion about this episode

User's avatar

Ready for more?