AI Safety Playbooks: Nvidia, Cisco, CrowdStrike Build Separate Frameworks
Nvidia, Cisco, and CrowdStrike are each developing their own AI safety playbooks to address growing security risks in modern AI systems. The three companies are working independently, with no joint guidelines or collaborative documents publicly reported.
Nvidia launched the Open Secure AI Alliance on July 27, 2026, to build open tools for responsible AI deployment and cybersecurity defense. The initiative includes the NVIDIA Labs Object-Oriented Agent project (NOOA), which targets emerging risks around AI agents. NOOA aims to secure autonomous software systems that can take actions without human approval.
Nvidia is also developing the Halos framework, focused on physical AI applications like robotics. When AI controls a robot arm or self-driving system, security failures have physical consequences.
Cisco published its Integrated AI Security and Safety Framework in December 2025, which identifies 19 attacker objectives and maps out over 150 attack techniques tied to recognized security standards including NIST and OWASP.