The cybersecurity defense has welcomed a new intelligent assistant, as Cisco officially open-sourced the small AI model family Antares on Tuesday. This model is specifically designed for software code repository investigations, accurately locating known security vulnerabilities in the code.

Compared to large AI models that often require significant investment, Antares demonstrates remarkable efficiency while maintaining high accuracy. Testing shows it can scan 500 code repositories in just 15 minutes, with a cost of less than $1.

Focused on local privacy protection, learning from hackers' approaches to accurately search

A key advantage is that this small model supports running locally within enterprises, completely eliminating the need to upload sensitive source code to third-party service providers. Its training logic resembles that of an experienced security investigator, enabling it to autonomously trace clues within massive codebases.

To address potential risks of misuse, Cisco has collaborated with relevant organizations to establish a download verification mechanism. Currently, the Antares-350M and 1B versions are available on open-source platforms for verified defenders.

Promoting the open-source security ecosystem, industry collaboration strengthens network defenses

The more powerful 3B version is not yet publicly released; Cisco plans to integrate it directly into its own security product line. Cisco is exploring the establishment of an industry alliance to further expand cooperation in the field of open-source AI security tools.

As multiple companies have launched related tools, industrial collaboration in the open-source security field is gradually deepening. This efficient and low-cost AI protection tool will build a more solid defense for cybersecurity in the digital age.