Why Aikido’s New AI Model Could Change the Way Companies Protect Sensitive Code

Belgian cybersecurity company Aikido has launched a new AI model designed to help businesses detect and address software security risks without necessarily sending sensitive code to external AI providers.

The launch highlights a growing shift in the technology industry: companies are increasingly looking for AI tools that can operate inside their own infrastructure, particularly when the information being analysed is highly confidential.

Aikido’s new model, called Altar, is an open-weight AI model customized for cybersecurity applications. According to Reuters, the company plans to make it available for local deployment and use it in solutions provided to customers including Belgian bank Belfius.

What is Aikido’s Altar AI model?

Altar is designed specifically around cybersecurity-related tasks rather than being a general-purpose chatbot.

The model is a compressed and customized version of Z.AI’s open model GLM-5.3. Aikido says it can be deployed locally, allowing organizations to run cybersecurity AI within their own environment.

That distinction is important because cybersecurity teams often work with information that companies do not want to send outside their networks.

This can include:

• Proprietary source code
• Software architecture
• Vulnerability information
• Internal applications
• Security configurations
• Sensitive business systems

With local deployment, companies can potentially use AI-assisted security analysis while keeping that information within their own infrastructure.

Why does local AI matter for cybersecurity?

The biggest issue is not simply artificial intelligence. It is where the data goes.

When a company uses a cloud-based AI service to analyse sensitive information, that information has to interact with an external service under the provider's specific architecture and policies.

For ordinary tasks, this may not always be a major concern.

Cybersecurity is different.

Source code can reveal how an application works, while vulnerability information can identify weaknesses that attackers could potentially exploit.

A locally deployed AI model gives organizations another option: keep the model and the sensitive analysis closer to the systems being protected.

This is particularly relevant for industries such as banking, healthcare, government and financial services, where data protection and regulatory requirements can be significant.

AI is becoming part of the cybersecurity battle

There is another reason companies are investing in defensive AI.

Cybercriminals are also using artificial intelligence to accelerate parts of the attack process.

Reuters reported that concerns about criminals using AI to identify and exploit software vulnerabilities are contributing to demand for defensive AI tools that can operate locally.

This creates an interesting cybersecurity race.

Attackers can use automation to search for weaknesses faster. Defenders increasingly want automation to identify and respond to those weaknesses faster.

Traditional cybersecurity teams may have to investigate enormous amounts of information manually. AI can potentially help identify patterns, analyse code and prioritize security issues more quickly.

However, AI does not eliminate the need for cybersecurity professionals. Its effectiveness still depends on the quality of the model, the underlying security data and how organizations validate its findings.

Why banks could be especially interested

Financial institutions are among the organizations that have the most to lose from a major cybersecurity incident.

Banks operate large digital infrastructures containing highly sensitive customer and financial information.

Aikido has already been working with Belgian bank Belfius, with the company's new AI model expected to be integrated into solutions for customers including the bank.

Earlier this year, Aikido and Belfius also announced an on-premise cybersecurity deployment involving AI agents designed to continuously test systems for vulnerabilities. VLAIO has separately described Aikido's approach as a move toward continuous rather than periodic security testing.

This points toward a broader change in cybersecurity: instead of checking systems occasionally and producing a report, companies are increasingly exploring continuous automated testing.

The bigger trend: AI that stays inside the company

Aikido's announcement fits into a wider movement toward local, private and sovereign AI infrastructure.

Businesses have already been experimenting with AI assistants that run within controlled environments because of concerns surrounding privacy, compliance and intellectual property.

Cybersecurity could become an especially important application for this model.

Imagine a company developing a new financial application.

Its security team wants an AI system to inspect the application's code for potential vulnerabilities.

Sending that code to an external service may create additional privacy, compliance or intellectual-property considerations.

A locally deployed cybersecurity model provides another architecture:

Company code → Local AI model → Security analysis → Internal security team

The sensitive material can remain within the organization's infrastructure rather than being sent to an external AI service.

Aikido's growing position in Europe's cybersecurity market

The launch also comes as Aikido expands its position in the European cybersecurity sector.

Wavestone's 2026 Belgian Cybersecurity Startup Radar identifies Aikido as the country's cybersecurity unicorn and says 31 of the 40 companies assessed in its ecosystem integrate AI into their products or services.

Wavestone also reported that Aikido accounted for €74 million of the €88.6 million raised by six Belgian cybersecurity companies covered in its funding analysis as of May 2026.

Reuters reported that Aikido reached a $1 billion valuation in January 2026.

What could happen next?

The significance of Altar may extend beyond one company's product launch.

If businesses increasingly want AI to analyse sensitive systems without moving that information to external providers, locally deployable models could become an important part of enterprise cybersecurity.

The challenge will be balancing three things:

AI capability + data privacy + security reliability

A cybersecurity AI system must not only identify potential weaknesses. Companies also need confidence that its analysis is accurate, that sensitive information remains protected and that security teams can verify its recommendations.

That means the future of cybersecurity AI may not simply be about having the biggest model.

It could increasingly be about having the right model running in the right place.

Key Takeaway

Aikido's Altar launch reflects a broader shift in enterprise technology. As companies become more cautious about sending sensitive code and security information to external AI platforms, locally deployable AI could offer a way to combine automated cybersecurity analysis with greater control over sensitive data.

For businesses, the question is gradually moving from "Can AI help protect our systems?" to "How can we use AI without giving up control of the information it needs to analyse?"

Source: Reuters / Global Banking & Finance Review; additional industry context from Wavestone and VLAIO.