About Breachwire

Breachwire is a research desk. We document real security breaches in AI/LLM applications, web applications, and progressive web apps, and turn each one into something you can act on: a root cause, the scan that would have caught it, the control that prevents it, and a red→green regression test that proves the fix.

How we verify

Every named-company breach cites a public disclosure — a CVE record, a vendor advisory, a government alert, or reputable security reporting. A vulnerability with no single named victim is labelled a documented vulnerability class and cites the taxonomy (OWASP, CWE, MITRE ATLAS), never a company. Nothing here is unsourced.

Our method

A bug is not real until it is reproduced. A fix is not done until a red→green check proves it. A defense is not trusted until the attack is shown to work without it and fail with it. We apply that discipline to security research so our reports are testable, not just narrated.

Frameworks we map to

• OWASP Top 10 for Web Applications (2025)
• OWASP Top 10 for LLM Applications (2025)
• OWASP ASVS 5.0
• NIST AI Risk Management Framework
• NIST Secure Software Development Framework (SP 800-218 / 218A)
• MITRE ATLAS and MITRE ATT&CK