5 Sept 2026·3 min read
Dependency Drift: How AI Tools Expand Your Attack Surface
AI-assisted development quietly introduces new dependencies, version bumps, and build hooks. Here is how to govern dependency changes before they reach production.
Blog
Analysis and field notes from the Cyblox team.
1-9 of 20 posts
4 Sept 2026·3 min read
Bot Defense KPIs and Governance: How to Prove It Worked
If you cannot measure outcomes and explain enforcement decisions, bot defense becomes a permanent tuning exercise. Here is the KPI model that holds up.
29 Aug 2026·3 min read
Prompt-to-Code Leakage: Preventing Secrets in Generated PRs
How sensitive data, internal context, and credentials leak from prompts into AI-generated code — and where to catch them before merge.
28 Aug 2026·3 min read
Scraping Is Competitive Intelligence at Scale
Scraping is no longer just copying pages. It is pricing intelligence, inventory monitoring, and model training — executed continuously and quietly.
23 Aug 2026·5 min read
For Cyber Attacks, the Human Is Not in the Loop Anymore. But for Cyber Defense, the Human Is Still Stuck in the Loop.
AI is removing humans from the cyber attack loop faster than it is removing them from the cyber defense loop. That asymmetry is becoming one of the biggest security problems.
22 Aug 2026·4 min read
The AI-Generated PR Review Checklist (Node/TS + Python)
A practical checklist for reviewing AI-assisted pull requests without slowing teams down: fast checks, deep checks, and what to automate pre-merge.
21 Aug 2026·3 min read
Credential Stuffing Is a Workflow Attack, Not a Login Attack
Defenders protect the login box. Attackers optimize the entire journey: login, reset, MFA fatigue, session reuse, and downstream abuse.
15 Aug 2026·4 min read
The Context Problem: When Safe Actions Hide Risky Diffs
Most risky outcomes in AI-assisted development aren’t triggered by obviously dangerous actions. They hide behind routine steps whose meaning changes with the diff.
14 Aug 2026·3 min read
The CAPTCHA Tax: Why Friction Is Not Control
CAPTCHA is easy to deploy and easy to justify. But in high-value channels it often becomes a recurring UX tax while adaptive automation keeps moving.