Nvidia and 36 Giants Form AI Security Alliance Without OpenAI: $35 Million Crypto Heist Heightens Concerns

Led by Nvidia, 37 tech giants have formed an open-source defense alliance on the grounds that closed AI models hinder forensic analysis.

The cards are being reshuffled in the tech world; Nvidia and 36 companies, including Microsoft, IBM, and SpaceXAI, have united under the Open Secure AI Alliance. This move comes especially after last week’s data breach on the Hugging Face platform and increasingly complex attacks targeting cryptocurrency protocols. Unlike closed models, the alliance aims to develop transparent and auditable security tools.

Alliance members have begun offering tangible tools to strengthen defensive walls. Nvidia released the NOOA framework, which simplifies testing AI behaviors; Microsoft shared the MDASH system, which detects errors. Meanwhile, SpaceXAI supported the process by open-sourcing its coding-focused Grok Build tool. These tools will serve as a digital shield against the malicious use of AI.

AI Shield for Crypto Protocols

Last week, the theft of more than 35 million dollars in assets from AFX, Verus, and the Bitcoin (BTC) scaling network B² revealed the scale of the danger in the crypto world. Attackers exploited secure control mechanisms using multi-step and complex methods in which AI is becoming increasingly proficient. Unlike traditional cyberattacks, this new line of defense is vital for developers since funds drained from smart contracts are irreversible.

A striking detail was the exclusion of giants like OpenAI, Google, and Anthropic from this alliance. The fact that closed models cannot distinguish between attackers and defenders during an attack and hinder investigations has increased the importance of open-source models. Nvidia emphasizes that when defenders cannot examine the AI within their own infrastructure, the speed of intervention is limited. This situation creates a new front in the tech world’s “open vs. closed model” battle.