Cybersecurity is a constantly evolving landscape, with threats accelerating at machine speed. In response to this, open-source large language models (LLMs) have emerged as a crucial infrastructure for startups and cybersecurity leaders to develop cost-effective defenses against rapidly evolving threats.
The benefits of open-source LLMs include faster time-to-market, greater adaptability, and lower costs. Recently, at the RSAC 2025 conference, industry giants like Cisco, Meta, and ProjectDiscovery announced new open-source LLMs and a community-driven attack surface innovation that is reshaping the future of cybersecurity.
One of the key takeaways from RSAC 2025 is the shift towards strengthening infrastructure at scale through open-source LLMs. The vision of creating a unified, open-source LLM and infrastructure to combat complex threats is becoming a reality, as highlighted by industry leaders at the conference.
Cisco’s Foundation AI group, built on Meta’s Llama 3.1 architecture, introduced the Foundation-sec-8B LLM. This cybersecurity-specific model, with 8 billion parameters, was purpose-built and meticulously trained on a cybersecurity-specific dataset. By open-sourcing the model, Cisco is enabling collaboration among cybersecurity providers to create more unified defenses.
The Foundation-sec-8B model’s training regimen included vulnerability databases, threat behavior mappings, threat intelligence reports, red-team playbooks, incident summaries, and compliance guidelines. This tailored approach ensures enhanced accuracy, contextual understanding, and quicker threat response capabilities.
Meta also expanded its AI Defenders Suite with Llama Guard 4 and LlamaFirewall, offering tools for compliance monitoring and AI agent protection. Additionally, ProjectDiscovery’s Nuclei, an open-source vulnerability scanner, won the “Most Innovative Startup” award at RSAC 2025, showcasing the power of open-source cybersecurity tools.
The success of these open-source initiatives aligns with Gartner’s recommendation for organizations to embrace open-source AI and cybersecurity tools. By fostering collaboration and expertise across boundaries, companies like Cisco, Meta, and ProjectDiscovery are leading the way towards more effective and affordable cybersecurity defenses.
In conclusion, the future of cybersecurity lies in openness, collaboration, and specialized domain expertise. As cybersecurity leaders embrace bespoke security models, better security efficacy can be achieved at a fraction of the cost, ultimately enhancing the overall cybersecurity landscape.
