RSA Conference 2025 produced a firehose of product launches across its 650-exhibitor expo. Here are the ten AI and data security announcements that mattered most, distilled for busy security leaders.

1. Cisco launches Foundation AI, the first open-source security reasoning model.

Cisco released Foundation-sec-8B, an 8-billion-parameter LLM pre-trained on curated cybersecurity data and published under Apache 2.0 on Hugging Face. Built to accelerate SOC workflows, threat defense, and code review, it marks the first major open-source AI model designed exclusively for security applications. Cisco also unveiled AI Supply Chain controls to detect malware in AI model files.

2. CrowdStrike unveils GenAI Data Leak Prevention.

CrowdStrike’s unified Falcon Data Protection suite introduced “Similarity Detection DNA” — technology that recognizes sensitive content even when users modify it before uploading to AI tools. Companion features include shadow AI detection, AI model scanning for trojanized code, and cloud-native runtime data protection using eBPF.

3. Palo Alto Networks introduces Prisma AIRS.

A comprehensive AI security platform covering the full enterprise AI stack: model scanning, posture management, AI red teaming, runtime security, and agent protection. The same week, Palo Alto announced its intent to acquire Protect AI.

4. Microsoft previews 11 Security Copilot agents.

Six Microsoft-built and five partner-built autonomous agents entered preview, handling tasks from phishing triage to conditional access optimization. Microsoft also expanded Purview with browser-based DLP controls preventing sensitive data from reaching GenAI applications.

5. Google Cloud debuts the Agentic SOC vision.

Google unveiled its roadmap for AI-powered security operations, including alert triage and malware analysis agents built on Gemini, integrated with Mandiant intelligence and VirusTotal.

6. NVIDIA ships DOCA Argus for AI factory security.

A new DOCA software framework running on BlueField networking hardware provides runtime cybersecurity for AI workloads, operating independently of host CPUs.

7. Meta releases LlamaFirewall and Llama Guard 4.

Meta open-sourced LlamaFirewall, a real-time security framework incorporating PromptGuard 2, Agent Alignment Checks, and CodeShield. Llama Guard 4 adds multimodal content classification.

8. ProjectDiscovery wins the 20th Innovation Sandbox.

The open-source vulnerability management platform, powered by the Nuclei engine with 11,000+ community-contributed detection templates, took the top prize. Each finalist received a $5 million investment from Crosspoint Capital.

9. IBM launches autonomous threat operations.

IBM’s Autonomous Threat Operations Machine (ATOM) brings agentic AI to threat triage, investigation, and remediation. A companion X-Force Predictive Threat Intelligence agent delivers industry-specific threat forecasting.

10. Cryptographers’ Panel sounds post-quantum alarm.

Adi Shamir warned the industry was ignoring quantum computing’s threat to current encryption. Raluca Ada Popa urged immediate adoption of hybrid cryptographic approaches. NetApp announced embedded post-quantum cryptography across its storage portfolio.