CYGNVS launches platform to manage AI system failures

Cyber resilience firm CYGNVS has released a dedicated platform for managing crises triggered by an organization's own artificial intelligence systems, addressing a category of operational risk that existing incident response tools were not built to handle.

The AI Incident Command Center extends CYGNVS' out-of-band incident platform to cover failures in production AI models and autonomous agents. Those failures include biased outputs, hallucinations, data leakage, and agents that pursue objectives in unintended or harmful ways, according to details first reported by SiliconANGLE.

Isolation by design

The platform's core architecture isolates incident response from corporate networks and the AI systems under investigation. This prevents a malfunctioning or compromised AI from detecting, influencing, or manipulating the response process — applying the same containment principles security teams use during ransomware incidents.

CYGNVS built the system around a proprietary dataset of more than 20,000 major incidents drawn from the insurance industry. The platform integrates with AI deployments to surface failure signals across applications, models, and agents, then activates matched playbooks and escalates response into its isolated environment.

Why it matters

AI incidents are no longer edge cases. The OECD AI Incidents and Hazards Monitor recorded 596 AI incidents in January 2026 alone, a 200 percent increase year-over-year. Gartner research cited by CYGNVS found that 61 percent of senior professionals report seeing AI agent automation deployed through approved enterprise software, while 59 percent report evidence or strong suspicion of unsanctioned AI agents running outside governed pathways. Most organizations lack infrastructure to coordinate response when those agents fail, defaulting to email and messaging channels that a malfunctioning AI may be able to access.

Four-stage response framework

The platform covers four stages: Prepare, Practice, Respond, and Report. Organizations enter an incident with a playbook tailored to their industry, geography, and incident type. They can run tabletop exercises to build readiness, coordinate live response in a single logged environment, and file regulatory notifications using prebuilt templates.

Those templates cover requirements across 56 binding laws and 47 frameworks globally, including the EU AI Act, the California AI Act, New York Local Law 144, the Colorado AI Act, and FDA AI and machine learning guidelines.

Matt Honea, chief information security officer at Hippocratic AI, said the platform addresses a timing problem. "The time to deploy AI incident response is alongside the AI project rollout, not afterwards and playing catchup," Honea stated. "AI incident readiness requires playbooks, tabletop exercises, coordinated response and incident reporting, mirroring exactly what cybersecurity teams have built over the last decade."

Founded in 2020, CYGNVS raised a $55 million Series A round led by Andreessen Horowitz in January 2023. The company reports that more than 3,000 organizations use its platform and that customers run more than 50 major incidents per week on it. The AI Incident Command Center is now generally available.

SiliconANGLE first reported the launch.