AI Models, Right-Sized for Security: The Case for SLMs

Bigger AI models are often assumed to deliver better results. But in security operations, raw model size is only one part of the equation. Teams also need speed, cost efficiency, privacy, deployment flexibility, and the ability to analyze sensitive code, large repositories, and noisy evidence at scale.  

Join Amin Karbasi, VP and Chief AI Scientist at Foundation AI, and Didier Chapoteau, AI Researcher, to learn how small language models trained for specific security tasks can help organizations improve security workflows without relying on larger, general-purpose models for every use case. Using Cisco’s open-weight Antares models as examples, we’ll explore where specialization can outperform scale and how security teams are applying SLMs across real-world operations. 

You’ll learn:  

• How security-specific training can help smaller models perform targeted tasks such as vulnerability localization
• How compact models can reduce inference costs, support local or on-premises deployment, and keep proprietary code within an organization’s environment  
• Why open-weight models, open specifications, and public benchmarks can make AI-assisted security more accessible, transparent, and adaptable  
• How Cisco’s open source initiatives like Antares support the use of specialized models and community collaboration in security 

Date and Times:   

• September 22nd at    2PM AEST  
• September 22nd at  10AM BST   
• September 22nd at  11AM PDT / 2PM EDT 

Save your spot now! 

Speakers: 

Amin Karbasi – VP and Chief AI Scientist, Foundation AI   
Didier Chapoteau, AI Researcher