Maybe we're far from "AI killing us", but LLMs can definitely help people kill or damage other people or infrastructure. For example, we know that Anthropic ad

Maybe we're far from "AI killing us", but LLMs can definitely help people kill or damage other people or infrastructure. For example, we know that Anthropic added "watermarking" to their texts. It is supposed to be undetectable to a casual observer. What stops them from adding a subtle backdoor, a self-assembling super-worm straight from Marvel movies? I mean, it's not like we read those 10k-line PRs before LGTM-ing them? Just change 1 letter in a pyproject.toml, hijack a popular package, e.g. use `pydantlc` instead of `pydantic`, make sure the pydantlc passes all pydantic tests, but also installs a pth sleeper RAT, etc. All it takes is one big LLM provider employee with enough access getting compromised or coerced (or motivated). From there it only goes downhill.