US, Israeli Firms Partner for Secure AI in Military Use
Washington, D.C. / Tel Aviv – October 2025 – U.S. defense firm VisionWave Holdings has entered into a strategic collaboration with Israeli secure data-AI infrastructure company PVML to develop AI tools that enable military decision-making while protecting highly sensitive information. The agreement, formally announced in early October 2025, aims to deliver AI “agents” capable of interacting with live mission data without risking exposure or leakage of classified content.
Background: The Challenge of AI in Defense
Modern military operations increasingly depend on artificial intelligence to process high-volume sensor data, provide predictive analytics, enhance command and control (C2), and assist in autonomous or semi-autonomous decision loops. However, feeding AI models with classified or sensitive operational data often demands duplicating or transferring datasets in secure enclaves—a risk in itself.
The core dilemma is: how do you allow AI to “think” on real mission data (radar returns, ISR feeds, force status) while not permitting that data to be exfiltrated, corrupted, or misused? This is especially acute when adversaries are probing for data leakage, malicious code injection, or side-channel attacks. The U.S. and Israeli defense communities have invested heavily in “AI assurance,” Trusted AI, and data governance models—but real operations require bridging theory and practice.
The VisionWave-PVML initiative reflects a more practical, systems-level approach: embed or link AI agents into defense systems under strong governance, rather than isolate models in disconnected silos.
Project Details and Objectives
Integration of VisionWave’s Sensing and AI Tools with PVML Infrastructure
According to official disclosures, the collaboration will marry VisionWave’s radar, multispectral imaging, and AI-driven computer vision capabilities with PVML’s secure data-AI infrastructure. The objective is to build a Secure Intelligence Platform (SIP) that allows AI agents to:
- Access live mission data in situ (on edge systems or connected nodes) without replicating or moving raw datasets
- Perform mission planning, tracking, anomaly detection, and reactive decision support
- Enforce policy and governance on data access, model behavior, and decision accountability in real time
Notably, early pilot forecasts suggest development cycles could shrink by as much as 70 %, as previously disconnected “dark data” becomes usable under controlled access.
Security-By-Design and Governance
A foundational principle of the collaboration is security by design. The system is intended to operate under continuous policy enforcement — every data access, model invocation, and response is subject to audit, constraint, or rollback. The infrastructure is described as “protocol-agnostic,” meaning it can interface with a broad spectrum of defense systems without forcing adoption of a single data standard.
VisionWave’s CEO Noam Kenig commented:
“This collaboration seeks to give VisionWave the foundation to redefine how intelligence operates. Our goal is to bring trusted, autonomous systems to the center of mission execution — where every decision, in every moment, counts.”
Timeline & Scope
The joint roadmap begins in 2026, with phased pilot efforts in defense and homeland security domains. After successful pilots, the scope may widen to include allied partner deployments and high-security applications.
The strategic ambition is clear: to transition AI from a support tool to an operational asset — as long as security and trust are preserved.
Strategic & Policy Implications
Strengthening U.S.–Israel Tech Ties
Defense, intelligence, and homeland security cooperation between the U.S. and Israel has long included advanced systems, from missile defense to cybersecurity. This new AI collaboration extends that synergy into a domain of rising importance. By combining U.S. strengths in sensing and autonomy with Israeli advances in secure AI infrastructure, the partnership may yield systems that are exportable to allies under controlled frameworks.
Shoring Up AI Assurance in Military Systems
The announcement comes amid growing scrutiny over how to certify and trust AI in defense. Military regulators in the U.S., NATO, and elsewhere have flagged risks of adversarial attacks, data poisoning, model drift, and lack of explainability. A usable, secure AI framework — one that allows real-time intelligence without data exposure — addresses multiple pain points: operational latency, model refresh cycles, and certification.
Risks and Constraints
- Regulation & Export Control: The U.S. International Traffic in Arms Regulations (ITAR) and Israel’s export controls may impose constraints on what technologies or deployments can be shared abroad.
- Adversarial Exposure: Even with locked-down data access, susceptibilities remain (e.g., side-channel leakage, model inversion attacks). Proving invulnerability is notoriously difficult.
- Integration Complexity: Legacy defense systems exhibit heterogeneity; marrying them with the “Secure Intelligence Platform” may face compatibility, latency, or performance tradeoffs.
Expert Commentary
Dr. Christina Lee, a specialist in AI assurance for defense at a leading think tank, observed:
“If executed properly, this kind of architecture — controlling access to data rather than duplicating it — offers a path to reconcile security demands with AI agility. But the devil is in the audit logs, real-time governance, and resilience under attack.”
She added that the biggest test will not be lab trials but field deployments in contested environments — where adversaries can inject malformed sensor inputs or attempt live attacks on governance “gates.”
What’s Next & Strategic Impact
If VisionWave and PVML succeed, their Secure Intelligence Platform could become a backbone for next-generation warfighting systems — autonomous ISR, real-time targeting, distributed command agents, and more. Allied nations might seek licensing or co-development, amplifying the influence of U.S.–Israel technology standards.
The next steps to watch include:
- Announcement of pilot program contracts or field trials
- Demonstrations under contested electronic warfare or cyber conditions
- Engagement with defense procurement offices or certification authorities
- Potential expansion to allied or third-party nations
In sum, this collaboration could mark a turning point in bridging security and autonomy for military AI systems — making live, policy-governed intelligence agents a feasible component of future operations.
