NVIDIA Launches AI Safety Team Focused on Open-Weight Models and Cybersecurity

NVIDIA is forming a new AI Safety and Cybersecurity Engineering team aimed at enhancing its AI safety capabilities. Last month, the company first disclosed the team's existence through job postings, announcing plans to hire roles including founding technical leads, security research engineers, evaluation engineers, and senior managers. The team will be responsible for conducting safety assessments of AI agents prior to deployment and developing AI-powered tools to automatically patch software vulnerabilities. NVIDIA stated that open-weight models—those that publicly release trained parameters while keeping training data and source code closed—alongside technical transparency and broad scientific scrutiny, form the foundation for the United States to maintain its leadership in AI and strengthen its cybersecurity defenses. In recent months, NVIDIA has repeatedly voiced support for open-weight AI models, contrasting with the closed-model approaches adopted by companies like OpenAI and Anthropic.