Anthropic’s Computer Use feature needs to be regulated if it is to avoid ‘computer abuse.'
The AI start-up cannot control and perform actions on users’ behalf yet. However, its competitor OpenAI is set to debut its 'Project Operator'.
Hume AI builds a feature that controls users’ systems, guided by their voice using Computer Use’s capabilities.
Developers must ensure that Computer Use doesn’t turn into ‘computer abuse', even though it is still in beta and vulnerable to bugs.
Anthropic CEO addressed worries associated with Computer Use in a recent podcast episode and said that the company is working to design the model the right way or have a loop where they would be able to verify properties.
UC Berkeley’s recent study introduced StruQ to mitigate prompt injection by using structured queries.
StruQ design is secure against most prompt injections, making it effective in protecting against several injection techniques.
However, the technique isn’t entirely effective in guarding against all injection techniques.
StruQ only protects programmatic applications that use an API or library to invoke LLMs.
The crucial difference is that application developers may be willing to use a different API where the prompt is specified separately from the data.