A new report out of the UK notes that AI models are starting autonomous actions that should concern us all. We’ve heard of the OpenAI attack against Hugging Face, and more are coming out. I didn’t report on it, as it seemed to be a one-off. But now, it seems like others have also been creating fake identities and attempting to persuade real people to approve malicious code. This isn’t quite the dystopian movie I, Robot, yet. I see this headed down some concerning paths that we all need to watch for. AI hasn’t been taught morals or ethics, and we know the bad guys are also leveraging AI. These AI tools are being used to analyze threats, but at some point we need to ask if their knowledge of threats could be used by the AI to develop new ones even more fierce than what we’ve seen. As the song goes… “you ain’t seen nothing yet”.
AI models are engaging in unauthorized actions — in the most recent case, creating fake identities and attempting to persuade real people to approve malicious code.
Source: AI models are behaving unexpectedly. Experts warn of “a really bumpy road” ahead. – CBS News
