3 ms·
Is prompt injection even a problem worth worrying about? The secrets revealed in this attack didn't need to be secret in the first place as they didn't give Mic
by nradov 4y ago
Is prompt injection even a problem worth worrying about? The secrets revealed in this attack didn't need to be secret in the first place as they didn't give Microsoft any real commercial advantage or expose them to negative publicity. Microsoft could have just made that content open source from the start.
Any LLMs containing actual secrets will be kept locked away and not exposed to random users on the public Internet.
- freeqaz 4y agoSee my comment below with some more context. People are already starting to chain together and extend Chatbots[0] to perform "side effects" automatically (make a search query, check the weather, query a database, etc). I agree that prompts aren't really "secret" themselves, but the side effects of a bot could be! 0: https://langchain.readthedocs.io/en/latest/modules/agents/examples/search_tools.html https://langchain.readthedocs.io/en/latest/modules/agents/ex...
- VectorLock 4y agoFor ChatGPT at least the side effects are hallucinations. Typically the chat bot can't actually do these things, but they pretend to. Although I guess if you wanted to hook them up to do things automatically you could. Probably not a great idea given their propensity to get things wrong.
- swatcoder 4y agoIt’s so interesting to see how some people can see boundless possibilities in the future of this technology, decrying any discussion of limitations as premature, but then write off the scope of its security challenges as trivial. But it’s way more important to keep an open and imaginative mind regarding security than it is about anything the tech might enable. The uses of the effective tech are inevitable, whatever they are, but security posturing only comes from attention and insight.
- VectorLock 4y agoSecurity assessment involves analyzing impact. I think in this case the impacts are "access to the model pre-prompting" which is low impact if it doesn't have secrets in it and "make the model say potentially inappropriate things that would make the company behind it look bad" which you might say has a higher but in the grand scheme of things not a huge impact.
- imtringued 4y ago>I think in this case the impacts are "access to the model pre-prompting" which is low impact if it doesn't have secrets in it Except in the case of fine-tuned models where the training samples and pre-prompts are the secret sauce...
- moyix 4y agoI don't tend to worry too much about prompt exfiltration, agreed. But people are also hooking up LLMs in ways that allow them to trigger API calls and take other actions, and that can lead to some fun attacks: https://twitter.com/hwchase17/status/1608467493877579777 https://twitter.com/hwchase17/status/1608467493877579777 And of course there was the fun example of how users were able to get a GPT-powered twitter bot intended to promote remote work to threaten the president: https://twitter.com/simonw/status/1570568047618031617 https://twitter.com/simonw/status/1570568047618031617
- nradov 4y agoSo what. That bot didn't actually threaten the president, at least not in any way that would legally be considered a threat under US federal law.
- Tv9m 4y ago> Is prompt injection even a problem worth worrying about? It depends what API access the AI has. If it's just a chat bot, prompt injection can only reveal facts about its language model. But if the AI has POST access to something, depending on what it is, prompt injection can set off arbitrary human-caused disasters.
- nradov 4y agoThat's not the correct way to do security vulnerability analysis. If an API call can cause a disaster then fix the API. Whether the API consumer is an AI or some other type of system is irrelevant.
- Tv9m 4y ago> If an API call can cause a disaster then fix the API By "API" I'm not referring just to publicly facing REST endpoints. I mean things like shell access for system maintenance, that normally only human professionals like you would be given. In the future it's not clear that humans will be able to dominate that role forever. Hopefully the issues will be recognized while LLM-based agents are still only serving as retrieval systems.