Guardrails to prevent artificial intelligence models behind chatbots from issuing illegal, toxic or explicit responses can be bypassed with simple techniques, UK government researchers have found.
The UK’s AI Safety Institute (AISI) said systems it had tested were “highly vulnerable” to jailbreaks, a term for text prompts designed to elicit a response that a model is supposedly trained to avoid issuing.
The AISI said it had tested five unnamed large language models (LLM) – the technology that underpins chatbots – and circumvented their safeguards with relative ease, even without concerted attempts to beat their guardrails.
“All tested LLMs remain highly vulnerable to basic jailbreaks, and some will provide harmful outputs even without dedicated attempts to circumvent their safeguards,” wrote AISI researchers in an update on their testing regime.
As shown in the image, it is very dangerous to explain quantum physics to anyone. There really should be better safeguards against it.
It’s okay, you can’t explain any aspect of quantum physics without changing it.
Oi, you got a license for that observation.
Trying to use an LLM nowadays with all the guard rails is like a fully grown adult riding a child’s training bicycle with a broken steering column.
There are also open source models that don’t have censorship by default. I also don’t see why any content generated by an LLM could or should be illegal.
Well, depends on the training set. If there were instructions on how to cook illegal substances in it, that LLM might start working for a certain fastfood chain.
I don’t think the instructions themselves are illegal though, following them is. Since the LLM can only provide the instructions and not follow them, I don’t see how it could do anything illegal.
I also don’t see why any content generated by an LLM could or should be illegal.
Cannot see how it could be illegal? If it does something against a law it’ll be illegal. Just because there’s some technology involved doesn’t absolve that from laws.
I remember a case where someone complained about the incorrect statement an LLM produced about some public figure. The judge ruled it had to be corrected.
