New Chinese AI model raises concerns over hacking risks
Kyiv • UNN
Chinese company Z.ai has introduced the GLM-5.2 artificial intelligence model, which experts believe could simplify hacking attacks. It has an open-source code and costs half as much as equivalents from OpenAI and Anthropic.
Chinese company Z.ai has introduced a new artificial intelligence model GLM-5.2, which has raised concerns among cybersecurity experts. Specialists believe it could significantly simplify the execution of hacker attacks, reports UNN citing Axios.
Details
According to estimates by Graphistry and Semgrep, GLM-5.2 performs at the level of leading models from OpenAI and Anthropic in vulnerability detection and cyber threat analysis. At the same time, its use costs approximately half as much.
GLM-5.2 has an open-source code, so it can be run locally, security settings can be modified, and it can be adapted for various tasks. According to experts, this significantly facilitates the use of the model by cybercriminals. Discussions on ways to bypass its protections have already appeared on Russian-language forums.
Why this matters
Unlike commercial services, open-source models provide virtually no ability to control their use. This allows attackers to create their own tools for phishing attacks, vulnerability discovery, and other malicious activities without the risk of being blocked by the developer.
SpaceX investor predicts imminent emergence of AI data centers in orbit25.06.26, 02:38 • 3762 views