Cisco AI

Watch Cisco Cloud Control in action. Discover the unified platform built for humans and AI agents to run critical IT infrastructure together, enabling customers and partners to build their own apps and agents in natural language, and extending to third-party tools.

The brightest signals come from the smallest sources. For years, AI has chased bigger models. We've been exploring a different question: What if the smartest security model isn't the biggest one? Tomorrow, something new arrives from @Cisco. Stay tuned.

Most cybersecurity benchmarks measure what models remember. Real security work measures how they reason. That's why we introduced two new benchmarks in FAITH (Cisco Foundation AI's Testing Hub): • CTI-Reasoning: Multi-hop reasoning over MITRE CAPEC & CWE • CWE-Prediction: 2025 CVEs and recent GitHub Security Advisories beyond most training cutoffs Open source and challenging even for today's frontier models. Read how they work from @fdtn_ai scientist @kimimajd: cisco-foundation-ai.github.io/blogs/new-cybe…

Every great AI experience deserves a great defense. Something exciting is teeing up with @Cisco and @USGA…

How do today's AI models adapt stories for different cultures? Last week, @Cisco researchers presented new work at @aclmeeting 2026 showing that AI-generated stories often rely on shared narrative templates, changing only a small set of culturally specific details rather than the story itself. Read the paper to learn more: t.co/IYbM9vybzE

More products. One connected platform. Cisco Cloud Control brings capabilities across networking, security, compute, observability, and collaboration into one operational environment - giving teams and AI agents shared context and control.

Last week at @icmlconf 2026, researchers from @Cisco and collaborating institutions presented two papers covering LLM safety and agentic AI security. Congrats to our authors, collaborators, and research teams. Read the papers: 📄 Distill to Detect: Exposing Stealth Biases in LLMs through Cartridge Distillation - t.co/nU9eAMtFIU 📄 CrossAnchor: Image-Anchored Text Optimization Exposes Blind Spots in Multi-line Defenses of Agentic Systems - openreview.net/pdf?id=kMOXzRN…

Earlier this week at ACL 2026, our joint work with the @awscloud GenAI team on benchmarking RAG was presented at the GEM Workshop. The findings from this research have been instrumental in making our Agentic RAG systems work at enterprise scale. Congratulations to the @Cisco researchers and engineers behind this work and to everyone who contributed to bringing it from research into production. Read more: arxiv.org/html/2606.2565…

Do you actually know where your AI models come from? Our open source Model Provenance Kit is like a DNA test for AI models. Released in April, the Model Provenance Kit generates a rich "fingerprint" using metadata signals, tokenizer similarity, and weight-level identity signals to determine whether AI models share a common origin or training lineage. Read more: blogs.cisco.com/ai/model-prove…

AI agents can reason, call tools, access data, and take action at machine speed—expanding the AI attack surface. Join @Cisco’s Emile Antone June 30 at 11 a.m. PT to explore agentic AI risk and how Cisco AI Defense can help teams validate and protect AI applications and agents. Watch live: cloudsecurity.cisco.com/webinar-learn-…

Build, deploy, and scale AI agents with confidence. See how Cisco AI Defense makes this possible. Follow @ciscoai for more on securing AI applications and agents.

Room & Board’s network infrastructure team is just three people responsible for connectivity across the business. Mark Rodrigue shared his perspective with @BizTechMagazine on the practical value of Cisco Cloud Control for his team: more cohesive management, fewer disconnected platforms, broader visibility and better context for AI-assisted troubleshooting. Learn more: biztechmagazine.com/media/video/ci…

We are thrilled to share how @SoftBank uses our Foundation AI team’s Foundation-Sec-1.1-8B-Instruct model for their SOC workflow. By categorize software into 17 policy-relevant groups, the model helps drive the workflow with policy verification and response actions. Check out our blog post in the comment.

The future of IT isn't more dashboards. It's intelligent systems that help teams move faster, reduce complexity, and stay in control. Learn how from Cisco President & CPO @jpatel41 via @NetworkWorld. cs.co/6017BDP8YH

ICYMI 🌟 In this episode on N is for Networking, Robert Barton, an AI Distinguished Engineer at Cisco Systems, explores how agentic AI is transforming network operations and what it means for your career. He joins Ethan and Holly to help snap the artificial intelligence puzzle piece into the networking picture. Together they break down the AI trifecta: domain knowledge, system fluency, and model understanding. Listen here or wherever you get your podcasts: cs.co/6013BDnERH

For decades, security came at the expense of productivity. Jeetu Patel believes that trade-off is finally coming to an end. As President and Chief Product Officer at Cisco, @jpatel41 is focused on making trust the foundation of the AI era. Autonomous agents are beginning to act on behalf of businesses, and the cost of a single error can have far-reaching consequences. As that risk grows, so does the need for a new kind of defense. Security has to evolve and operate at machine speed. His conviction runs deeper: he believes AI’s true promise lies not in aggregating what we already know, but in generating insights that do not yet exist. These insights could transform how we treat critical diseases, manage risk, and defend nations. @tech_mahindra and @Cisco have been building toward that future together across 5G, firewall modernization, and next-generation security. Our collaboration continues, grounded in the balance between human judgment and AI-driven efficiency. Read the latest edition of #TheBigThinkers: partners.wsj.com/tech-mahindra/…

Manual prompt engineering doesn’t scale. Meet FAPO (Fully Automated Prompt Optimization): an AI coding driven system, supported by both Claude Code and Codex, that autonomously optimizes LLM pipelines from baseline prompts to production-ready accuracy. It beat GEPA on 15 of 18 benchmarks (+33.8pp on retrieval-heavy tasks). Open source now: cisco-foundation-ai.github.io/blogs/fully-au…

Join Amy Chang, Head of AI Threat Intelligence and Security Research, as she speaks at the UNIDIR Global Conference on AI, Security and Ethics this Friday, June 19, for a panel discussion on the convergence of AI and disarmament. Watch live here: youtube.com/live/Cf8-MO7DY…

You shipped a new model checkpoint. Did it regress on your eval benchmarks? If your results live in scattered JSON and CSV files, answering that means writing another one-off merge script. So we stopped treating eval metrics like log files and started treating them like data. FAITH now writes every run to a structured BigQuery table. Regressions become a GROUP BY. Provenance becomes a column. Ingestion is one line. Our Foundation AI Research Scientist, @kimimajd, wrote about how we built it: cisco-foundation-ai.github.io/blogs/scaling-…

Infrastructure that thinks, adapts, and acts—on its own. That's what Cisco Cloud Control delivers for the AI era. See how in the Keynote Deep Dive with @DJSampath and team. blogs.cisco.com/ai/cisco-cloud…