Research - Escape - Application Security & Offensive Security Blog

Research

AI vs AI: How Cascade exploited an AI agent in production

We pointed Cascade, Escape's AI pentesting engine, at a production AI agent. The agent had a guardrail built to catch prompt injection. Cascade got past it on the second try. The trick wasn't a smarter payload, just a different pretext, and the agent handed over its entire system prompt: the tools it could call, the rules it followed, and the session details attached to the conversation. That alone gives an attacker what they need to go straight at the agent's tools instead of guessing at them

Yacine Souam

August 13, 2026 4 min read

Research

Escape Research team found a PII disclosure in Keycloak. It's now CVE-2026-17059.

Keycloak filters its main users list so a restricted admin sees nothing. The endpoint that lists a role's members skips that filter and hands the same account everyone's email and name. Escape research found it, reported it, and it's now tracked as CVE-2026-17059.

Enzo Mongin

July 31, 2026 10 min read

Research

Why does the rise of Agentic AI force security teams to become research-aware?

Bruce Schneier : "security is a process, not a product". It’s not enough to audit your application once, every new feature introduces new attack surfaces that must be taken into account when evaluating the security of the application as a whole. For example, a company rolling out a new Agentic AI process without adequate guardrails might expose themselves to MCP tool poisoning, an emerging class of attacks where malicious actors leverage the trust placed in a tool metadata to steer th

Arnaud Fanthomme

July 24, 2026 8 min read

New exploitable flaw found in Immich (BOLA)

DAST

How Escape DAST Bypassed Immich's Locked Folder By Finding a Missing Default

How the Escape DAST uncovered a missing default that lets any Immich user read locked-folder photos without the PIN.

Enzo Mongin

July 16, 2026 11 min read

AI pentesting

Modern AI-powered Pentesting Tools In-Depth benchmark

If you're evaluating AI for offensive security right now, you're weighing two questions at once. The first: why not skip the tooling and point a frontier model at your apps yourself? The second: among the AI pentesting tools you could actually buy, which one earns the budget? We built this benchmark to answer both. Cascade, Escape’s multi-agent AI pentesting engine, went up against a single frontier model (Claude Opus 4.8) driven directly at the same targets. The only variable is the harness ar

Antoine Carossio

June 30, 2026 15 min read

AI pentesting

Two Critical Vulnerabilities, One AI Pentester: How Cascade Found an Unauthenticated RCE and Walked Around the WAF

TL;DR We pointed Cascade, Escape's AI pentesting solution, at a single Spring + JSP customer portal. It came back with two findings that are typically difficult for traditional Dynamic Application Security Testing (DAST) scanners to detect: * Unauthenticated RCE via SpEL injection. A ref request parameter was dropped, unsanitized, into a spring:eval expression inside a global JSP template. Because the JSP engine substitutes the value before SpEL parses it, an attacker controls the expression

Karim Rustom

June 30, 2026 11 min read

Agentic Pentesting

How Escape AI Pentesting Exploited SSRF in LiteLLM

At Escape, we routinely test the AI infrastructure that teams deploy inside their cloud environments. LLM gateways, RAG pipelines, model proxies: these are services that make outbound HTTP requests by design, which makes them natural targets for SSRF. When we looked at LiteLLM, we found three confirmed SSRF sinks, a security gate that was specifically built to prevent this exact bug class, and a bypass that means the gate itself is broken. The organization has confirmed to our research team th

Yacine Souam

May 1, 2026 9 min read

Application Security

Benchmarking AI Pentesting Tools: A Practical Comparison

We benchmarked 4 AI pentesting tools: Escape, Shannon, Strix, and PentAGI against a modern vulnerable application. Learn more about their detection rates, false positive rates, and scanning speed.

Gwendal Mognier

April 30, 2026 12 min read

Research

Methodology: How we discovered over 2k high-impact vulnerabilities in apps built with vibe coding platforms

Hey there, With Halloween around the corner, what’s scarier for organizations than vulnerabilities in their web applications? And it's even scarier when the development of these applications is in the hands of users not familiar with security practices. This year, the Escape research team has focused on a growing area of concern: vibe-coded applications. Purpose-built platforms like Lovable.dev, Base44.com, and Create.xyz have democratized application development, enabling non-developers to d

Nohé Hinniger-Foray, Gwendal Mognier, Alexandra Charikova

October 29, 2025 10 min read

Research

Fortune 1000 at risk: How we discovered 30k exposed APIs & 100k API vulnerabilities in the world’s largest organizations

Hey there! Next week, social media will be flooded with Thanksgiving feasts (and Black Friday deals). But before you dive into the holiday shopping madness, the Escape team has prepared a special treat for you—not a Black Friday deal on Escape, of course, but rather some impressive findings on API security risks among Fortune 1000 and CAC 40 organizations. And trust us, the findings aren’t just “spicy” like your Thanksgiving stuffing—they’re quite scary... Using a combination of advanced subd

Alexandra Charikova, Maxence Lecanu, Quentin Lieumont, Gabriel Marquet

November 20, 2024 6 min read

Research

The State of Public APIs 2023

tl;dr we scanned 6056+ public APIs on the internet with our in-house feedback driven exploration tech and ranked them using security, performance, reliability, and design criteria. We decided to analyze the resulting data and produce a full featured report: The State of Public APIs 2023 Why build this report? 🤔 Two weeks ago, we released apirank.dev, the result of extensive research on public APIs on the internet, conducted at an unprecedented scale using our feedback-driven API exploration

Escape - Offensive Security

March 23, 2023 2 min read

Announcement

APIRank.dev: we scanned and ranked 6031+ public APIs on the internet

tl;dr we scanned 6031+ public APIs on the internet with our in-house feedback driven exploration tech and ranked them using security, performance, reliability, and design criteria. The results are public on APIrank.dev. You can also request to index your own API for free and see how it compares to others. Why APIrank? 🤔 During a YC meetup, I spoke with fellow founder Andrew Kim from Finex. He told me about their struggle to select the right API providers in his past companies. It's true tha

Tristan Kalos

March 10, 2023 3 min read