AI Guardrails Are Hobbling Offensive Cybersecurity Research: Why Defenders Need the Weapons AI Companies Won't Let Them Use
Strict AI guardrails from Anthropic and OpenAI are blocking legitimate security researchers from finding and verifying vulnerabilities, paradoxically weakening cybersecurity defenses worldwide.
The AI Guardrail Dilemma
In July 2026, TechCrunch highlighted a growing tension in cybersecurity. AI guardrails — the safety measures that Anthropic, OpenAI, and other companies install to prevent malicious use of their models — are paradoxically impeding legitimate cybersecurity researchers who need these very tools to find and confirm vulnerabilities.
Leading Researchers Speak Out
Mark Dowd, a renowned security researcher who has spent decades finding and selling zero-day exploits, criticized AI companies for making "arbitrary decisions about what is safe in security." Chris Anley, chief scientist at NCC Group, emphasized that asking an AI model to attempt exploiting a bug is a key step in confirming whether it's a real vulnerability worthy of fixing. "If a guardrail prompts the model to refuse to answer outright, the guardrail hurts defenders."
The Hammer That Cannot Be Unmade
Anley offered a powerful metaphor: AI tools are "like a hammer. You can't build a house without a hammer. It's definitely a tool, but it's also irreducibly a weapon as well." The same instruction — "fix this code" — can be both a defensive mechanism and an offensive roadmap. When researchers hit roadblocks, they often fall back on open-source models that come with no guardrails at all, highlighting the futility of containment.
Vetted Access Programs
OpenAI offers a "Trusted Access for Cyber" program and Anthropic a "Cyber Verification Program." Paolo Stagno, CTO of Crowdfense, argued these programs "essentially treat customers like children who need babysitting." The issue: Anthropic's Mythos had US export controls placed on it in June 2026 after claims its guardrails could be bypassed — controls that have since been partially lifted — showing the complex interplay between AI companies, governments, and the security community.
Implications for Georgian Tech
For Georgian technology companies, this debate carries concrete implications. As businesses increasingly rely on AI for everything from code review to penetration testing, dependence on closed models with arbitrary guardrails creates a fragile foundation. Open-source models offer flexibility and control, especially in security-critical applications.