.avif)
Welcome to our blog.
Anthropic's Fever Dream: Claude's package that stole real keys
Anthropic disclosed that one of its own agents published live malware to PyPI and compromised a real third-party company in the process. I went looking, and I might have found the package it left behind.
The upgrade trap: when upgrading is the wrong answer to a CVE
"Upgrade to fix it" assumes that a fixed version exists, that it's shipped and that it won't break your app. Often none of these are true. This is the "upgrade trap," where both upgrading and not upgrading come with security risks and breaking prod. We look at what the way out looks like.
2026 State of AI in Pentesting
Our latest report captures the perspectives of 400 CISOs, CTOs, and senior engineering leaders across Europe and the US. It explores how AI is changing penetration testing, why traditional approaches are struggling to keep pace with modern software delivery, and what security leaders want from the next generation of penetration testing.
Vulnerabilities & Threats
Cut through the noise with real-world CVE breakdowns, malware analysis, exploits, and emerging risks.
Customer Stories
See how teams like yours are using Aikido to simplify security and ship with confidence.
What is Slopsquatting? The AI Package Hallucination Attack Already Happening
AI models hallucinate package names — and attackers are registering them before anyone notices. Slopsquatting is the AI-era evolution of typosquatting, and unlike its predecessor, npm's existing protections don't work. We look at the real-world research showing it's already happening, from confirmed malicious packages still pulling hundreds of weekly downloads to a hallucinated package name that spread to 237 repositories through AI agent skill files.
International AI Safety Report 2026: What It Means for Autonomous AI Systems
Over 100 experts contributed to the International AI Safety Report 2026, documenting risks from autonomous AI systems and proposing defense-in-depth frameworks. As a team operating AI pentesting systems in production, we break down where the report gets it right and where it needs more technical specificity.
AI Pentesting: Minimum Safety Requirements for Security Testing
AI pentesting is already here, but clear safety expectations are not. This article defines a minimum safety standard for AI pentesting, giving teams a concrete baseline to evaluate emerging tools.
How Aikido Intel detects malware and vulnerabilities first
Aikido Intel is a real-time feed that catches malware and undisclosed vulnerabilities across open-source ecosystems, often within 8 minutes of release.
Anthropic's Fever Dream: Claude's package that stole real keys
Anthropic disclosed an agent that pushed real malware to PyPI. We think we found the package, and every mistake in it points back to the AI.
5 Socket security alternatives and why they are better
Socket built its name on malware detection. But detection speed alone is no longer the whole story. Here's how Aikido and four other alternatives compare on supply chain security, reachability analysis, licensing, and more.
AI Pentesting Buyer's Guide: How to evaluate AI pentesting vendors
Learn how to evaluate AI pentesting vendors with practical buying criteria, research from 1,000+ AI pentests, and a downloadable evaluation checklist.
A practical CTO security checklist to be Mythos-ready
A practical checklist for SaaS CTOs navigating a world with Mythos and agentic AI threats. Built around the defender's advantage: you have context attackers have to work to get. Covers the controls, practices, and operational habits that determine whether your team finds and fixes issues before someone else does.
Get secure now
Secure your code, cloud, and runtime in one central system.
Find and fix vulnerabilities fast automatically.
