A new, freely available AI model from China is changing the rules in cybersecurity, which is the protection of digital systems. This is the GLM-5.2 model from Zhipu AI, a Chinese AI company. This model can find security vulnerabilities almost as well as Anthropic's secret US model Mythos.
This development is a real power shift in the digital world. Previously, the US led in specialized AI systems for cybersecurity. Now, a similarly powerful tool is freely entering the market, which anyone can use. This makes powerful defense tools accessible to everyone. At the same time, it also opens up new opportunities for attackers. This race has just begun.
The Chinese AI company Zhipu AI has released its open-source model GLM-5.2. According to a test, this model finds security vulnerabilities almost as well as the US model Mythos. The US government keeps Mythos under wraps. GLM-5.2 is a so-called open-weight model. This means its blueprint is public, and anyone can use and adapt it.
For you as an individual, your online security becomes more complex. On the one hand, new apps and services could better protect your devices. On the other hand, hackers using such open-source models could find new ways to access your data. You need to be even more vigilant about what software you use and what data you share. The line between secure and insecure is blurring more and more.
Companies face a major challenge. If attackers use AI models to find vulnerabilities in software and networks, companies must significantly improve their cyber defenses. This costs money and requires new strategies and trained personnel. Smaller companies that cannot afford expensive experts are particularly at risk. Attack tools are becoming cheaper and more accessible for everyone.
The free availability of GLM-5.2 also creates opportunities. Developers worldwide can use the model to build new security solutions. They can also improve existing systems. Open-source software can often be developed faster and more openly. Many people look at the code and find errors. This can lead to many new ideas in cyber defense. Ultimately, everyone could benefit from this.
The biggest risk is misuse. If such a powerful cybersecurity model is freely accessible, criminals or states with malicious intentions can also use it. They could use it to find vulnerabilities in important infrastructures or corporate networks. Such vulnerabilities would otherwise have remained undiscovered. It is almost impossible to control who uses this technology and for what purpose. This makes the global cyber landscape unpredictable.
