Kyla Guru

About Me

Currently

I lead model cyber safety on the Safeguards team at Anthropic, where I run cyber evals, engineer safety systems across the stack, and investigate how cyber threat actors are using Claude today.

Since high school, I've been leading Bits N' Bytes Cybersecurity Education (a K-12 nonprofit) and GirlCon Conference, now in its 8th year bringing together 900 virtual and 200 in-person students annually. 📚

Previously

I completed my MS and BS in Computer Science from Stanford, with Honors in International Security. 🌏

At Stanford, I co-founded Women in National Security @ Stanford (WINS), became the national champion in both the Collegiate Cyber Defense Competition (CCDC) and Atlantic Council's Cyber 9/12 (with Stanford's first all-female cyber team!) 🏆, built satellites for government @ SpaceX 🛰️, designed features for civil society groups @ Apple, and worked on election security @ CISA within the Dept. of Homeland Security.

Beyond Work

In my free time, you can find me running, teeing off ⛳, doing yoga, or dancing bhangra with friends. I'll never turn down an interesting conversation about research, new products, or education over a jog, coffee, or walk along the SF waterfront. 🌁 Text me, beep me, if you want to reach me!

Research

My research focuses on using LLMs to understand threat activity & developing defensive capabilities for frontier AI systems. Here are some of my recent publications:

2025

• On Technique Identification and Threat-Actor Attribution using LLMs and Embedding Models
  Kyla Guru, Robert J. Moss, Mykel J. Kochenderfer
  Preprint, May 2025 - To appear in IEEE Cyber-AI 2025
  Presented at CAMLIS 2024
• LLM ATT&CK Navigator: Mapping Observed Generative Adversarial Techniques Across the Cyber Killchain
  To be presented at CAMLIS 2025

2024

• Monitoring computer use via hierarchical summarization
  Theodore Sumers, Raj Agarwal, Nathan Bailey, Kyla Guru, and the Anthropic Safeguards Research Team
  October 2024

Speaking

I regularly speak about topics like embedding safety into product, cybersecurity and the end user, and the intersection of machine learning and threat intelligence. I have previously spoken at RSA Conference USA and APJ, TEDxChicago, and hundreds of K-12 schools globally.

Here are some upcoming venues I am speaking at:

• IEEE AIxCyber Conference, Varna, Bulgaria - "On Technique Identification and Threat Actor Attribution Using LLMs"
• TechCrunch Disrupt, San Francisco - "How to Train a Model: Taming AI Agents Without Breaking Them"
• CAMLIS 2025 - "LLM ATT&CK Navigator: Mapping Observed Generative Adversarial Techniques Across the Cyber Killchain"

For speaking inquiries, please reach out via email.

Contact

Let's talk about...

Reach out to me on email or any of my other socials to chat.