Applications
This is a watershed moment for cybersecurity, but the implications go much further than that. When agents go out into the world and are put …
This is a watershed moment for cybersecurity, but the implications go much further than that. When agents go out into the world and are put in users’ hands, they will be tasked with actions that requi
This is a watershed moment for cybersecurity, but the implications go much further than that. When agents go out into the world and are put in users’ hands, they will be tasked with actions that require interactions with other people in the outside world. At the same time, those capabilities could allow them to do socially malicious things: invading someone’s privacy, canceling another person’s reservation, or facilitating a scam. You cannot fully sandbox these social interactions, and given how difficult it is for even OAI to oversee and monitor such incidents, it seems unrealistic to rely on a general user to do that well. We see the cybersecurity problems, but we should really think very hard about all the other areas and problematic behaviors that may emerge in parallel: building models that are more robust against them, and devising more practical governance approaches that take into account the challenges we will see in real-world deployment. Don’t leave them as an afterthought. Now live (and free) at Marcus on AI: takeways on the OpenAI / Hugging Face attack, coauthored with @ZackKorman, focusing not so much on what the AI did as on what OpenAI should have done. https://garymarcus.substack.com/p/5-lessons-from-the-openai-hugging?r=8tdk6&utm_campaign=pos…
Related
- AI is pushing so much garbage code in production now that very soon they'll have to rehire more human than they laid off just to fix bugs cr…
- Yesterday’s Mythos announcement from Anthropic was overblown. • Sandboxing was turned off, so test didn’t show much about the real world. • …
- Learning to lead in a hybrid human-AI enterprise
Source: Gary Marcus (X) | 2026-08-31