Cisco to release ‘world’s first-ever’ AI-powered security reasoning model
Cisco Systems said on Monday it has raised the bar for cybersecurity professionals, announcing the upcoming release of the world’s first-ever AI security model capable of navigating multifaceted security problems with adaptive reasoning skills.
Angel Garcia via Getty Images.
Cisco Systems said on Monday it has raised the bar for cybersecurity professionals, announcing the upcoming release of the world’s first-ever AI security model capable of navigating multifaceted security problems with adaptive reasoning skills.
The announcement, made at the annual Cisco Live conference being held in San Diego from June 8th through June 12th, was one of many enterprise AI innovations touted by the global telecommunications network and software solutions conglomerate.
The new “Foundation-sec-8 b-reasoning” model is built upon Cisco’s first foundational cybersecurity model, the “Foundation-sec-8b,” released by the California-based company in April.
While both models are built using the Llama 3.1 8B framework – an 8-billion parameter, open-weight Large Language Model (LLM) purpose-built for security – the newest release “enables the kind of sophisticated analysis and decision-making required in security workflows,” Cisco said.
🚨 Just launched during #CiscoLive: Foundation-sec-8b-reasoning—the world’s first security reasoning model.
undefined Cisco Security (@CiscoSecure) June 10, 2025
Read more in the blog announcement: https://t.co/8EkWySsa7X pic.twitter.com/HY6zfXauLB
The cybersecurity models are said to be pre-trained on a dataset curated by the Cisco Foundation AI team, specifically “to capture the real-world knowledge and context security professionals work with every day,” and currently lacking in other general models.
This training knowledge encompasses everything from vulnerability databases, threat intelligence reports, and real-world incident summaries to security tooling documentation across cloud, identity, and infrastructure domains to compliance references and secure development practices, including industry staples such as the CVE database, MITRE ATT&CK, NIST Cybersecurity Frameworks, and OWASP Secure Coding Practices.
The additional skills generated in the new “Foundation-sec-8b-reasoning” model, according to a Cisco blog post released Monday, include:
- System and Configuration Analysis: Evaluate system settings and configurations to identify vulnerabilities and improve security posture.
- Adversary Behavior Mapping: Correlate threat intelligence data with attacker tactics to predict and understand adversary behavior.
- Threat Detection and Analysis: Analyze logs and traffic to identify malicious patterns and enhance threat-hunting accuracy.
- Access and Privilege Management: Assess permissions and roles to uncover over-privileged accounts and mitigate insider threats.
- Context Enrichment and Investigation: Provide contextual insights to streamline investigations and support faster incident response.
Besides reducing the fatigue experienced by many cybersecurity teams, Cisco said the open-weight model is expected to foster innovation, adaptability, and accelerate defense deployment.
It will also allow security teams to gain more clarity in complex threat environments and help companies keep sensitive data in compliance with regulatory bodies.
The Foundation-sec-8b-reasoning model is set to be released this summer, the company said.
In the meantime, folks can check out a casebook of how the latest enterprise security model can be applied to real-world workflows on GitHub. The original Foundation-sec-8b model is currently available for download on Hugging Face.