Home / Blog
Field notes from the engagements
Blog
What we find, how it is exploited, and what actually fixes it. Written by the people who ran the test.
Application Security
We checked 1,237 fixes. One in four was not fixed.
Between October 2021 and August 2026 our clients sent 1,237 fixes back to us and asked us to confirm they worked. 307 of them did not.That is more…
Read itApplication Security
What 890 security engagements actually find
6,048 findings across 700 tests. The most common are configuration and hygiene. The most dangerous are authorization - 133 high and critical, against 115 for XSS.
Read itAI and LLM Security
Somebody Wired the Darknet Into Your AI. What Could Go Wrong?
One MCP server packing 66 tools, feeding onion-site content straight into your model's context. The darknet is the most hostile input that exists, and a model cannot tell…
Read itAuthorization and Access Control
Configured Is Not Enforced
AWS shipped four condition keys for DynamoDB and did not evaluate them at runtime. Your policy sat there, looked excellent in code, and did nothing.
Read itAI and LLM Security
Every Week Someone Asks Me When AI Will Replace Pentesters
Datadog released a scanner that sends your code to an LLM instead of matching patterns. After 20 years in this field: the engine was never the problem. What…
Read itApplication Security
The Most Expensive Vulnerability Is a Token Nobody Rotated
No sophisticated SQL injection. No state actor with three APT teams. A developer pushed code with a secret in it, and a drug worth billions walked out.
Read itCloud Security
Cloud Ransomware: Soft-Delete and Versioning Are the Whole Story
One over-permissioned Service Principal leaked through an old configuration, and we could read, delete, replace and encrypt every blob. The difference between 'restored in five minutes' and 'we…
Read itEngagements and Process
Learn Mode and Enforce Mode Are Sold in the Same Breath. They Are Not the Same Thing.
An agent that scans, finds and proposes, with a human deciding, is where the field should go. An agent that writes and applies exploits on production, unattended, is…
Read itSupply Chain
One Click, and Every Private Repo You Can Reach Is Theirs
A link, inside the tool you use every day. Simulated keystrokes install an extension, which lifts the GitHub OAuth token - and that token is not scoped to…
Read itCloud Security
Attackers Are Running Their C2 on Your Cloud, and You Are Paying for It
No fixed IP to block. No suspicious infrastructure for the SIEM to catch. The traffic belongs to Amazon and the invoice arrives at the end of the month…
Read itTell us what the system does and what worries you.
If a penetration test is not what you need yet, we will say so.