←
Detectify4,900 posts
Detectify avatar

Detectify

@detectify

Scaling real-world hacker research with AI to help AppSec teams find and fix their most exploitable risks before attackers do.

◷ Joined November 2011▧ 1,443 Photos and videos
Detectify avatar
Detectify@detectify·2h

About 80% of findings from AI-driven SAST tools are noise, our customers tell us. Agentic Tooling gives your agents a way to check which ones are real. scworld.com/brief/detectify-…

Detectify avatar
Detectify@detectify·Nov 12

The payoff is fewer tokens, less time and findings you can check. One example he cites from Anthropic: presenting tools as a code API cut context from about 150,000 tokens to 2,000. Read the full piece: blog.detectify.com/best-prac…

Detectify avatar
Detectify@detectify·Nov 12

His rules for agent tools: • Wrap whole workflows • Type the inputs • Keep the output lean • Namespace them (recon_subdomains, oob_listen) • Let the tool write the log “Your agent is not impressed by ASCII art, and neither is Doug McIlroy.”

Detectify avatar
Detectify@detectify·Nov 12

Models give confident answers that are often wrong. In one study, just renaming a variable made GPT-4 reverse a correct answer 17% of the time. As Luke puts it, “a language model is a confidence machine.”

Detectify avatar
Detectify@detectify·Nov 12

The setup: small CLI tools do the deterministic work (subdomains, payloads, DNS callbacks), the model makes the judgement calls, and MCP connects the two.

Detectify avatar
Detectify@detectify·Nov 12

Doug McIlroy wrote the design rules for AI agents in 1978: “Make each program do one thing well.” New on our blog: @hakluke on building a hacking agent out of tiny, boring tools 🧵

Post attachment
Luke Stephens (hakluke) avatar
⟳ Detectify retweeted
Luke Stephens (hakluke)@hakluke·Sep 22

Back in 2021 I wrote a "how to hack APIs" blog for Detectify. It ended up being one of the most-read things I've written that year, but APIs have changed a lot since then: GraphQL, AI-generated endpoints, all of it. So I rewrote the whole thing for 2026. Check it out👇

Post attachment
Detectify avatar
Detectify@detectify·Jul 30

Unauthenticated access in AI agent platforms like Ruflo exposes API keys, logs, & execution tools in one go. Detectify co-founder Johan Edholm spoke with @DarkReading about the expanding attack surface of AI infrastructure. 🔗 darkreading.com/cyber-risk/p…

Detectify avatar
Detectify@detectify·Jul 22

When adversaries weaponize LLMs, attack economics change overnight. @DarkReading details how threat actors turned jailbroken AI into a commercial offensive platform. Our CEO Rickard Carlsson breaks down why lowering the barrier to entry changes defense: darkreading.com/cyber-risk/h…

Detectify avatar
Detectify@detectify·Jul 7

.@DarkReading reports on "JadePuffer", the first fully agentic, LLM-driven ransomware attack that autonomously executed a network intrusion and extortion. Our co-founder Johan Edholm discusses how traditional incident response can't keep up with AI threats darkreading.com/cyberattacks…

Detectify avatar
Detectify@detectify·May 5

We are proud to recognize some of the top ethical hackers from our Crowdsource community. Thanks to their research submissions, our customers can access some of the best hackers in the world: labs.detectify.com/crowdsour…

Detectify avatar
Detectify@detectify·10 Nov 2025

We're now integrating real-world threat data into our AI Researcher Alfred to prioritize and generate protection against the most active CVEs, boosting security speed and relevance for all Detectify customers: blog.detectify.com/product-u…