Nvidia: AI security is an engineering problem across the agent stack

Private Safety Processing lets OpenAI review related API requests for safety without retaining prompts or responses, with records staying in the customer's environment. An automated safety agent reviews them, and OpenAI personnel cannot access the content.

The ungated site explains how security data pipelines cut SIEM ingestion costs while preserving detection coverage. It also covers preparing the data layer for AI-driven security operations.

Forrester Research founder and CEO George Colony says the AI boom may be entering a new phase, with custom chips challenging Nvidia's dominance and value shifting toward software. He warns the transition won't be smooth and companies must adapt to AI computing or risk being left behind.

David N. Miller discusses how higher cost of capital affects AI investment, pointing to flexible financing for hard-to-refinance assets and infrastructure beyond the AI boom. He also covers AI's long-term potential, concentration risks, and why selectivity matters for investors.

Filmed in London on October 14, 2026, Huang said "I don't build my own AI. Mostly I use a service called Perplexity." The clip was captured by Kim Isenberg during his Europe tour.

Microsoft AI CEO Mustafa Suleyman said that if the chance of runaway machine intelligence causing human extinction is really 10%, "we don't just slow down or pause. We absolutely have to stop." The clip comes from ControlAI, which frames such risk estimates as common among AI insiders and scientists.