I'm seeing a lot of people finding low quality ChatGPT integrations that are just proxied to the OpenAI API with a system prompt... It reminds me A LOT of malware in the 90s. Things are pretty harmless but are low key costing people money to deal with...

I think there is a lot about 90s cyber security that will need to be relearned all over again around these integrations and I'm just waiting for people to figure out these embedded chats could probably have their system prompt broken and used raw instead of paying for their own OpenAI API access...

#ai #openai #security #LLMs #api