The AI Security Podcast

Sign up to receive in your inbox: http://eepurl.com/i7RgRM

Tania Sadhani and Miranda R discussed various AI security topics, including critical CVEs affecting platforms like ChatGPT and Hugging Face, the potential for SharePoint Copilot in internal reconnaissance, and malicious npm packages targeting Cursor developers. They also covered the OASP Gen AI security initiative's Agent Name Service (ANS), the proposed AI.txt for controlling AI agent interactions, and Unit 42's framework for agentic AI attacks. Furthermore, Miranda highlighted security guidance from international agencies, Anthropic triggering ASL 3 for Claude Opus 4, Microsoft's AI red teaming playground, a significant data leak from an AI vendor, and the Israeli police's use of AI-hallucinated laws.

Ever wondered how security vulnerabilities are found in AI? Join us as we chat with Aditya, a Vulnerability Researcher at Mileva Security Labs!

Sign up to receive in your inbox: http://eepurl.com/i7RgRM

This week we note regular CVEs in AI libraries such as Nvidia TensorFlow and PyTorch. We discuss a novel prompt injection technique called "policy puppetry", along with malware dispersal through fake AI video generators and Meta's release of an open-source AI security tool set including Llama Firewall. We also covered Israel's experimental use of AI in warfare, Russia's AI-enabled drones in Ukraine, China's crackdown on AI misuse, Dreadnode's research on AI in red teaming, geolocation doxing via multimodal LLMs, safety research on autonomous vehicle attacks targeting inference time, Config Scan for analyzing malicious configurations on Hugging Face, Spotlight as a physical solution against deepfakes, and Reply Bench for benchmarking autonomous replication of LLM agents.

This week I'm joined by my friend Alberto, he has an incredible storied career - from data science, insurance, AI risk, advising Tesla.. check out his book here! 
https://www.amazon.com.au/Ethics-I-Facts-Fictions-Forecasts/dp/1636763650

We talk about Stanford Human-Centred AI's latest AI Index report, check it out here: https://hai.stanford.edu/ai-index/2025-ai-index-report

Did you know we have a fortnightly threat intel newsletter? We decided there was so much good research in there we have to talk about it here! We're joined by threat intel lead Miranda for this fortnight's biggest AI security news, coming out in this week's digest! http://eepurl.com/i7RgRM

My friend Shain joins me on the podcast to talk about his work with the OWASP MLSec Top 10 list and organisational guidance, as well as how he got here!

For info about the list and how to contribute, check out the link: https://owasp.org/www-project-machine-learning-security-top-10/ 

This week we talk about AI red teaming.. I can't quite believe we haven't talked about it already! We cover the origins of red teaming in the military, how red teaming is done in cyber security, and the fundamentals of AI red teaming. 

Resources:
https://academy.hackthebox.com/course/preview/introduction-to-red-teaming-ai
https://tryhackme.com/path/outline/redteaming

I didn't manage to get my act into gear to record a fresh episode between getting back from the UK and heading to New Zealand so I've pulled one from the vault.. it's a good one though! A year ago I was interviewed by Microsoft for their AI security series. Massive thanks to Microsoft for having me and Sarah Young for excellent interviewing. Check out the rest of the series, they have some really cool people on. This is also a good episode for anyone new to AI security, it covers most of the bases.

Link here: https://www.youtube.com/watch?v=sPQaJVnBSRQ&pp=ygUYaGFycmlldCBmYXJsb3cgbWljcm9zb2Z0

I didn't want to do a solo episode so instead I enlist free content support by speaking to the vendors at AI UK, run by the Alan Turing Institute, in London! Thanks very much to those people, links below here:
Datambit: https://datambit.com/
RAISE: https://raise-project.uk/

Contact us at [email protected]