📺 OpenAI–Hugging Face breach is a great example of how 'potent' these AI models are, says Cohere CEO
In a CMZ interview, Cohere co-founder and CEO Aiden Gomez discusses AI safety and cybersecurity after leaving Google Brain, where he co-authored the paper "Attention Is All You Need." The conversation covers the acquisition of cybersecurity startup Isle and the risks that AI models pose to critical systems.
■ Cybersecurity Risk and AI Models
- The reported OpenAI/Hugging Face breach as an example of model capability
- AI models described as potent cyber tools; skepticism toward sci-fi human-risk narratives
- Questions around rogue-agent narratives, including sandbox security and task setup
■ Isle Acquisition and Vulnerability Discovery
- Isle's profile: two years old, out of stealth one year earlier
- Over 500 CVEs found, with more than 200 high or critical
- Vulnerabilities in foundational internet software such as curl and OpenSSL
■ Critical Infrastructure and Financial Exposure
- Dependence of finance, telecommunications, power grids, and water treatment on software
- Economic consequences if these systems are disrupted
- Risk from both open and closed models
For viewers following AI safety, enterprise AI, and cybersecurity, this interview outlines why cyber capability is treated as a near-term risk. It provides context on how AI agents are being applied to vulnerability discovery and why critical infrastructure sectors are central to the discussion. Viewers can use it to frame their own risk assessments around AI-enabled cyber threats and defensive tooling.
この動画を紹介した CNBC Television の最新動画も、紹介付きで読めます。
📄 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。