AI
Belgium Just Shipped a Cybersecurity AI That Keeps Your Code at Home
Belgian cybersecurity firm Aikido released an open weight artificial intelligence model called Altar on Monday, built for cybersecurity work, and its sharpest selling point is a door that stays shut, companies can run AI security tooling on their own machines instead of sending sensitive code to outside providers.
The demand behind it is easy to read. Criminals are using AI to identify and exploit software vulnerabilities, and every security team wants defensive AI that runs locally, inside its own walls. A model you can deploy yourself turns the strongest objection to AI security tooling into a solved problem. The buyer keeps both, frontier capability and code that stays home.
The timing rides a bigger European current. Data sovereignty has moved from policy papers into procurement checklists, and banks in particular now ask where every byte goes before they sign. An open weight release answers that question in the most verifiable way possible, download it, inspect it, run it behind your own firewall.
Altar is a compressed and customized version of Z.AI's open model GLM-5.3, released openly for local deployments. The geometry of the stack is worth noticing, a Belgian firm tuning a Chinese open model for European banks, built entirely on openness. Compression matters here because a model small enough to run locally is a model a mid size company can actually adopt, and customization for security work means it speaks the language of vulnerabilities rather than small talk.
It is already going to work. Altar ships inside Aikido's own solutions for customers such as Belgian bank Belfius, which tells you the target buyer, institutions whose code is the crown jewels and whose regulators agree. Aikido reached a $1 billion valuation in January, 871.31 million euros, and this release is its bid for a bigger table, turning a security product company into a model provider.
The release also rides a broader shift in how serious buyers evaluate open weight models. A downloadable model can be benchmarked, red teamed and stress tested by the customer's own security team, an audit path a closed API keeps to itself. For regulated industries, that verifiability is the purchasing criterion that matters most.
Security teams get the template from this one, frontier scale compressed into something you can run yourself, audited by the people it protects. Your code gets frontier grade eyes on it and stays exactly where it belongs.
Quick answers
What is this story about?
Belgian cybersecurity firm Aikido released an open weight artificial intelligence model called Altar on Monday, built for cybersecurity work, and its sharpest selling point is a door that stays shut, companies can run AI security tooling on their own machines instead of sending sensitive code to outside providers.
Why does this story matter?
Security teams get the template from this one, frontier scale compressed into something you can run yourself, audited by the people it protects. Your code gets frontier grade eyes on it and stays exactly where it belongs.
Sources
New to crypto? Read the crypto glossary, browse frequent questions, read our story, or explore the story archive.