Aikido Launches Local Cybersecurity AI Model
Aikido has released an open-weight cybersecurity model designed to run locally so customers can keep sensitive source code inside their own environments.
Aikido is betting that cybersecurity teams want AI assistance without sending their most sensitive code to an external model provider.
What happened
The Belgian cybersecurity company released Altar, an open-weight model designed specifically for security work.
Altar can be deployed locally so customer source code does not need to leave the organisation's environment.
The model is based on a compressed and customised version of Z.AI's GLM-5.3 and will be integrated into Aikido's broader security products.
The company is also using the technology in deployments with customers including Belgian bank Belfius.
Why it matters
Security teams have unusually strong reasons to care about where AI workloads run.
Uploading proprietary source code to an external model creates privacy, compliance and intellectual-property concerns, particularly in regulated industries.
A locally deployed specialised model can reduce that trade-off while still automating parts of code and vulnerability analysis.
The bigger picture
Enterprise AI is fragmenting into two deployment models: powerful cloud-hosted systems and smaller specialised models that run inside customer-controlled environments.
Cybersecurity is likely to be one of the strongest markets for the second approach because data sensitivity is central to the product.
Aikido's launch reflects the growing overlap between AI sovereignty, open-weight models and security infrastructure.