Cisco's Chief AI Scientist Makes the Case for Owning AI Over Renting
shashib · x · 2026-08-12
At the Black Hat conference, Cisco's Chief AI Scientist Amin Karbasi shared the philosophy behind their vulnerability-scanning model Antares, arguing that enterprises should "own" rather than "rent" AI capabilities.
- Core Pain Point: Traditional code protection tools often require sending code externally, posing data security and compliance risks.
- Technical Approach: The team scaled the model down instead of up. Antares is small enough to run on a single GPU inside the customer's own environment without uploading code. It's trained to search a codebase like an investigator rather than answering questions like a chatbot.
- Cost Efficiency: This design cuts scan time from hours to minutes and drastically reduces costs. A 500-task sweep across 290 repositories costs only $0.82 with Antares, compared to $141 with GPT-5.5.
Related event: Cisco Open-Sources Antares for Cost-Effective Code Review(2 posts)→
More from coding & agent
- Claude + Obsidian + Loop Engineering: Building a Self-Running Vault — anirbanbandyo · 2026-08-12
- Embabel: Open-Source Agent Framework for JVM Hits 4.1k Stars — springrod · 2026-08-12
- Developer Building Fully Local Agentic Movie Studio, Plans to Open Source Everything — TheChuckTone · 2026-08-12
- Why AI Needs Memory Engineering: Graph Memory Beats Flat Vector Stores — anirbanbandyo · 2026-08-12
- Building an AI Content Agent: Why Chasing Throughput Is a Trap — Specialist-Band-7821 · 2026-08-12
- Bessemer Releases 63-Page Agentic AI Engineering Transformation Playbook — AccBalanced · 2026-08-12