AI Cyber Alliance Meet-up #3
Join us for another evening of cutting-edge insights into AI security and offensive operations in the SF Bay Area!
Date: May 13th
Food: Complimentary food and refreshments provided. Thanks for your continued support.
What to Expect
This month's meetup features elite speakers going deep on offensive AI security, coding agent internals, and red teaming at the frontier. Whether you're a student just starting your journey in security, a seasoned industry veteran, or anywhere in between, you're welcome here.
Program Highlights
6:00 PM to 6:15 PM | Networking & Refreshments
6:15 PM to 6:45 PM | Jackson Reed – Tollbooth: Inspecting and Manipulating Coding Agents
6:45 PM to 7:15 PM | Josiah Lashley – From Logs to Language: Using Encoder–Decoder Architectures to Translate Cyber Threats
7:15 PM to 7:45 PM | Open networking and community building
Featured Speakers
Jackson Reed is the founder of Barding Defense and a member of BT6, a top AI red team that frontier labs engage for pre-release security testing of their models and products. He spent four years rebuilding and leading a Fortune 50 red team, and before that built and led a Department of Defense red team through NSA certification. He works at the intersection of AI and offensive operations and writes about cybersecurity and AI at twelvetables.blog.
Talk: Tollbooth: Inspecting and Manipulating Coding Agents When coding agents run on YOUR machine, how do Anthropic and OpenAI protect their thinking tokens and system prompts? What user-controlled data can be injected into system prompts? Can you enroll yourself in hidden features? Jackson will walk through these questions using Tollbooth — an open-source tool that proxies all network traffic from a coding agent and its tool calls, enabling rich rulesets for modifying traffic on the fly. The talk covers Claude Code and Codex in detail: how they work, the attack surface their traffic exposes, and how OpenAI and Anthropic protect their models and agents differently.
Josiah Lashley is a founding AI engineer at DeepTempo and has been working in the field of cybersecurity with a focus on data science and statistical applications. His work centers on building intelligence pipelines that transform large-scale telemetry such as NetFlow, WAF logs, and cloud events into structured representations that power modern detection and response systems. He specializes in model-centric security architectures that integrate machine learning with operational SOC workflows.
Talk: From Logs to Language: Using Encoder–Decoder Architectures to Translate Cyber Threats Transformers were originally built for language translation, with encoders to understand and decoders to generate. Josiah applies that same paradigm to cybersecurity, where telemetry (IP flows, HTTP requests, auth events, behavioral sequences) has its own language, but analysts think in narratives, not raw logs.
The Encoder learns structured representations of attacker behavior from high-volume telemetry.
The Decoder translates those representations into human-readable context, timelines, and explanations.
The goal: turn signal into story. Attendees will leave with a new lens on detection engineering, AI-assisted SOC workflows, and how modern ML can bridge machine-scale data and human-scale reasoning.
Who Should Attend?
Everyone is welcome! This event is intentionally designed for all experience levels.
Why Attend?
Network with diverse professionals across the cybersecurity ecosystem
Learn from expert speakers at the frontier of AI security and offensive research
Connect with like-minded individuals passionate about security
Discover new opportunities and collaborations in the security space
Get hands-on insights into coding agent internals, attack surfaces, and AI red teaming
RSVP
Space is limited! Secure your spot today.
We look forward to seeing you there!