anthropics / anthropic-quickstarts

A collection of projects designed to help developers quickly get started with building deployable applications using the Anthropic API
MIT License
6.92k stars 1.03k forks source link

computer use with most secure way #195

Open musakurel opened 13 hours ago

musakurel commented 13 hours ago

Computer Use documentation have below warnings

  1. Use a dedicated virtual machine or container with minimal privileges to prevent direct system attacks or accidents.
  2. Avoid giving the model access to sensitive data, such as account login information, to prevent information theft.
  3. Limit internet access to an allowlist of domains to reduce exposure to malicious content.
  4. Ask a human to confirm decisions that may result in meaningful real-world consequences as well as any tasks requiring affirmative consent, such as accepting cookies, executing financial transactions, or agreeing to terms of service.

is there a way to use Computer Use with sensitive data for example login info? They say "Avoid giving the model access to sensitive data". But if we don't give sensitive data, can we still work with that data ?

JoshuaAIHAHA commented 8 hours ago

The main risk is if the AI goes to a website and is instructed to expose sensitive info it will. As well as data leaks and other factors. They mainly say that to warn you; that if you do and something does go wrong that its on you. I am just a user though and I have had zero issue with it in this regard. So you can do it at your own risk and it all depends what kind of work you are doing.