Cisco Introduces Antares: Small-Language Model Security Tools for Vulnerability Detection

Cisco has introduced Antares, a family of specialized small-language model (SLM) security tools aimed at making secure AI available to more organizations. The goal is to address the challenges posed by large-language models (LLMs), which can be exploited by bad actors and are often too expensive for smaller organizations.

Amin Karbasi, VP and chief AI scientist at Cisco, explained that Antares is designed to navigate code repositories efficiently and pinpoint vulnerabilities. This is particularly important for universities, nonprofits, public-sector institutions, and research organizations with smaller security teams, who face the same threats as large entities but often lack the resources to use frontier models.

Cisco has released two 'open weight' Antares models on Hugging Face, with a third more capable model approaching frontier-level performance. The company emphasizes that its goal is not to replace frontier models but to provide a specialized tool for vulnerability localization.