Who I Am
I’m Chris. I’m 32, based in Austin, TX. I run a YouTube channel called HaxDogma.
My background is in enterprise defensive security. For 15 years I worked in IT and security operations — writing IPS rules to detect and block real in-the-wild attacks, watching adversaries work in production systems, and building the detection logic to stop them.
I am not a penetration tester. My perspective comes from the defender’s seat: studying how adversaries actually behave, at scale, in real environments, and translating that understanding into systems that catch them.
What I’m Doing
I started looking at AI systems through the same lens I’d applied to network infrastructure for 15 years. The failure modes I found — the ways these systems can be manipulated, steered, and made confidently wrong — are reproducible, underreported, and present in systems being deployed at enterprise scale right now.
I decided to do something about it.
I’m publicly building a path to become CEO of Anthropic — the AI safety company that builds Claude. This is not a stunt. It is a 10-year plan, documented in the open, accountable to anyone watching.
The Declaration is the first video — a public commitment, on record. The Roadmap details every phase. The Research section documents the findings.
Published Work
Indirect Prompt Injection in Local LLMs — The first published research finding from this project. A systematic 36-test evaluation of indirect prompt injection against Llama 3.2 3B. The model followed attacker-embedded instructions in 77.8% of tests. Full methodology and data are published openly for scrutiny and reproduction.
Why Public
There’s a version of this where I spend five years quietly building credentials, then show up in a hiring manager’s inbox with a polished LinkedIn and no track record.
This version is more honest, more accountable, and more useful.
If I say I’m going to publish original AI security research and I don’t — you’ll notice. If I claim to understand AI safety and I demonstrate that I don’t — the community will correct me. That feedback loop is a feature.
Why Anthropic
Anthropic was founded on the belief that AI safety is existential. Their published work — the Responsible Scaling Policy, Constitutional AI, their interpretability research — is serious work done seriously. I have genuine respect for it.
What I believe the next generation of AI leadership needs — and what is currently underrepresented — is people who have spent their careers understanding how systems fail under adversarial conditions. That is a precise description of what I have done for 15 years.
Contact
If you have feedback on the research, find a methodology flaw, or want to discuss any of this — I want to hear it.