Altman Says Nvidia Safety Systems Not Foolproof, OpenAI to Launch Multiple New Models

Altman's framing of AI as a scientific rather than engineering problem is the tell: it implicitly lowers expectations for the kind of guaranteed safety that hardware-level guardrails promise, and it puts Nvidia's positioning under a spotlight it rarely faces from its largest customers. The Meta dismissal reads as competitive signalling rather than a technical claim, and the two together suggest OpenAI is trying to set the terms of the safety debate before its own model releases land. Whether that framing holds once those models ship is the open question.