AI Safety for Agents and Robots: Open Models, Identity, and Hardware Boundaries

1 件の動画 · 更新: 12時間前

このチャンネルのほかの紹介

Can Open Source Make AI Agents Safer? 📺 Can Open Source Make AI Agents Safer? ⏱ 16:40📅 2026/11/18 11:00

AI Safety for Agents and Robots: Open Models, Identity, and Hardware Boundaries

▼

A discussion on AI safety and agent security with Thomas Wolf and Amanda Grady, focused on why model-level guardrails alone are insufficient and what complementary controls are needed. The conversation examines open and proprietary models, agent identity, hardware-enforced policy, physical AI safety, and broader access to AI development.

■ Agent safety and controls
- Model-level guardrails versus external boundaries; NVIDIA Sentry and open models for cyber defense
■ Identity and runtime governance
- Agent-native identity; ServiceNow AI Control Tower, AI Passport, and hardware-layer enforcement; revocation and contextual constraints
■ Open-source security and scaling
- Security as an enabler for agent deployment; open versus closed model security and obscurity
■ Physical AI and access
- Robot safety and Microduck reference design; fine-tuning democratization; concentration of power in AI development

The discussion is suited to security leaders, AI engineers, product and policy teams evaluating layered AI controls, agent identity, and open-model strategies. Viewers gain a structured overview of the safety challenges and control approaches discussed and can use it to frame further assessment of deployment-specific safeguards.

この動画を紹介した NVIDIA の最新動画も、紹介付きで読めます。

いつも見ているチャンネルも、同じように紹介できます。

いつも見ているチャンネルを追加する

📄 このページの紹介文は AI が独自に生成したものであり、著作権をはじめとする他者の権利(商標権・名誉権・プライバシー等)を侵害しないよう配慮しています。動画の著作権は各作成者に帰属します。