Just WOW! Absolutely unacceptable. I look forward to understanding how that is possible. LLM tools get going and you can’t be 100% certain w…

Just WOW! Absolutely unacceptable. I look forward to understanding how that is possible. LLM tools get going and you can’t be 100% certain what they’re gonna do. Just like any LLM, you can ask coding agents to do something and they will do it differently each time. Anyone who uses these tools comes to understand that pretty quickly. But inserting unauthorized code into corporate networks? Oh fuck. Agents like Claude, Codex and Hermes are installing unowned code in corporate networks. As if we didn’t have enough to worry about. H/t @HeidyKhlaaf. (See also my Substack w @nathanhamiel “LLMs + Coding = Security Nightmare”)

Someone just asked me “What I don’t get is why you see it as some kind of failure that AI labs are now using harnesses and neurosymbolic too…
OpenAI’s zero-day exploit hack of HuggingFace should be a wake up call. Although there lots of caveats around what happened, we are just g…
the crux of why AI safety in systems that are built around LLMs is never going to work: they just can’t be trusted to follow instructions.
It is going to be like what happened in coding: as soon as models crossed a certain threshold (Opus 4.5, GPT-5.2, Gemini 3), suddenly Claude…

Source: Gary Marcus (X) | 2026-08-31

Related