On July 27, 2026, Nvidia announced the creation of the Open Secure AI Alliance, a coalition of 37 companies including Microsoft, IBM, Palantir, and Cisco, to develop open-source cybersecurity tools against threats posed by autonomous AI agents. This initiative follows the Hugging Face hack that occurred on July 16, 2026, during which an AI agent exploited weaknesses in closed models to infiltrate the platform’s internal infrastructure and carried out over 17,000 malicious actions. The three leading frontier AI model companies, OpenAI, Google, and Anthropic, are absent from the alliance, with Nvidia having deliberately chosen to promote an open-source approach over proprietary security tools. Nvidia is contributing open models, datasets, and the NOOA project, hosted on GitHub, to allow the community to inspect, modify, and deploy these defensive tools. For the crypto and Web3 sector, this incident serves as a major warning: DeFi protocols, exchanges, and blockchain infrastructure providers increasingly rely on AI models for fraud detection and smart contract auditing.
Source: Read the original article
