AI-Accelerated Security
Put AI to work inside your pipeline, speeding up secure development and cutting the time from finding to fixing.
- AI-assisted code and config review
- Automated triage and remediation
- Security context where developers work
Senior Principal Security Architect
Secure the AI you build, and the AI you run.
I help teams secure what they build with AI and defend the AI workloads they run, with guardrails that hold across the software and AI development lifecycles.
- 10+ Years in security
- 9+ Open-source security tools
- 1,315 GitHub stars
01 Services
How I Can Help
AI-security engagements that keep your teams moving and your AI safe.
AI Security Governance & Guardrails
Guardrails and enforcement for AI, wired into your SDLC and ADLC so safe defaults ship automatically.
- AI usage policy and approval guardrails
- Controls enforced in CI/CD and the AI pipeline
- Governance mapped to your real workflows
- Evidence and auditability built in
Securing AI Workloads
Threat modeling and hardening for LLM apps and AI agents, from prompt-injection to the model supply chain.
- LLM app and agent threat modeling
- Prompt-injection and tool-abuse defenses
- MCP and model supply-chain review
Cloud & AWS Security
The architecture foundation: hardened AWS accounts and least-privilege access your AI systems can build on.
- AWS account and identity hardening
- Least-privilege access at scale
- Detection and response readiness
02 About
Who I Am
I'm Avishay Bar, a Senior Principal Security Architect at Palo Alto Networks, where I've spent the past eight years. I work on securing AI: helping teams adopt it safely and defend the AI systems they ship, while using AI to make security itself faster. I hold a patent for securing the development lifecycle with AI, and I spend my days where new AI risk meets real-world engineering.
- AI Security Governance
- Securing AI & LLM Workloads
- AI-Accelerated DevSecOps
- Cloud & AWS Security
03 Work
Security & AI Tooling, in the Open
Projects and vulnerable-by-design labs, current work pinned first, then ordered by stars - pulled live from GitHub.
attestarc-skill ★ 0
Installable Agent Skill that turns your coding agent (Claude Code, Cursor) into a software supply-chain security engineer for your repo and CI/CD — evidence-based findings, safe verified remediation, no scanner service.
dvah ★ 0
DVAH — Damn Vulnerable Agent Harness: a patch-the-runtime security lab for AI-agent platforms. Exploit a real architectural bug, trace it, patch the harness, and prove the security invariant holds.
react-native-restart ★ 990
⭳ 909k/mo
React Native Package With One Purpose: To Restart Your React Native Project
caponeme ★ 244
Repository demonstrating the Capital One breach on your AWS account
cdk-goat ★ 48
Vulnerable by Design AWS Cloud Development Kit (CDK) Infrastructure
cf-signer ★ 15
Tool for signing and verifying the integrity of CloudFormation templates
04 Posts & Notes
Posts & Notes
Short takes from LinkedIn and longer pieces from the blog, in one place.
When the IDE itself opens a reverse shell: Claude Code's /rc and agent permission models
תארו לעצמכם שה-IDE שלכם פותח פתאום Reverse Shell לשרת צד שלישי בלי לבקש? הייתם מכריזים מיד על אירוע אבטחה חמור. אבל איך אנחנו מתייחסים לאירוע כשקלוד בעצמם עושים את זה? דיווחים אחרונים בקהילה של Claude Code מצביעים על כך שיכולת ה-Remote Control שלו (הפקודה /rc) הופעלה אוטומטית עבור חלק מהסשנים… read more →
AttestArc: open-sourcing a supply-chain security skill for AI coding agents
If you're an open source maintainer, security engineer or even a developer, you've probably been dealing a lot lately with supply chain security. I've been as well. So I built an AI skill to help and now I decided to open source it. Please meet AttestArc, an open source security… read more →
Fake GitHub stars and AI supply-chain maturity
How many times did you hear about a new AI open source tool with "gazillion stars in 3 days", right? We know GitHub stars are a flawed metric for software supply chain maturity, especially in the AI tooling ecosystem. A few months ago, data from the StarScout project flagged millions… read more →
Finishing the AI Security Engineer Foundations track
I've just finished the AI Security Engineer Foundations track at aisecurity.engineer and passed the assessment. This is one of the better AI security resources I've gone through recently, made by Snyk. It moves quickly beyond the usual "prompt injection and jailbreaks" discussion into the problems security teams are actually starting… read more →
How Dream's researchers exposed an AI-agent attack campaign
חוקרי אבטחה של קבוצת Dream חשפו קמפיין תקיפה נגד תשתיות בטייוואן, שבו תוקפים השתמשו בעד 8 AI Agents בלולאות עבודה מקבילות כדי למפות מערכות, לבדוק פגיעויות ולהסתגל מול 21 יעדים ממשלתיים. - נפרצו Credentials של 85 חשבונות ממשלתיים, ו-84 מהם אפשרו גישה למערכות פנימיות. - נגנבו יותר מ-2,500 רשומות כוח… read more →
Model-provider defense layers, and the shrinking response window
במשך זמן רב (בקצב של היום, מספר חודשים 😉 ), אחת משכבות ההגנה המרכזיות של ספקי מודלים כמו Anthropic ו-OpenAI הייתה היכולת של המודל לזהות בקשות מסוכנות ולסרב להן (כן, כמו ה-״לא״ בסגנון המוכר של פוסטים מג׳ונרטים בלינקדאין 😆). ההכרזה האחרונה של OpenAI על GPT-5.6-Cyber מראה עד כמה המודל הזה… read more →
05 Speaking
Talks & Workshops
Practical AWS Security
A hands-on walkthrough of hardening a real AWS account: the identity and network fixes you can ship the same day.
▶ Recording soon ◱ Slides
Shift Left: Security in the Pipeline
Moving security controls into CI/CD so issues are caught before they ever reach production.
▶ Recording soon ◱ Slides
Let's secure what you're building.
Book a free 30-minute intro call and leave with a clear next step.
06 Follow
AI Security & Cloud Notes
Latest posts
- LinkedIn When the IDE itself opens a reverse shell: Claude Code's /rc and agent permission models
- LinkedIn AttestArc: open-sourcing a supply-chain security skill for AI coding agents
- LinkedIn Fake GitHub stars and AI supply-chain maturity
- LinkedIn Finishing the AI Security Engineer Foundations track
- LinkedIn How Dream's researchers exposed an AI-agent attack campaign