What You Need to Know About AI Tools and Your Family's Safety

Microsoft researchers spent a year testing advanced AI systems to understand their security weaknesses. These are not simple chatbots, but AI assistants that can take actions on your behalf, like booking appointments or managing tasks. The researchers discovered seven new ways these systems can fail or be manipulated, including risks from the underlying technology supply chain and situations where hackers can redirect the AI's goals to serve their purposes instead of yours. This affects anyone using advanced AI assistants, especially those that can take actions in the real world beyond just answering questions.

If you use AI tools that connect to your email, calendar, files, or can make purchases or reservations for you, there is a risk that these systems could be manipulated or compromised. The concern is greatest with AI agents that have permission to act on your behalf without asking for confirmation each time.

Here is how to use AI tools more safely right now. First, review which AI tools and assistants you have given access to your accounts and what permissions they have. Second, whenever possible, require the AI to ask for your approval before taking any action, rather than acting automatically. Third, do not give AI assistants access to sensitive information like financial accounts, medical records, or confidential work documents unless absolutely necessary. Fourth, treat AI-generated suggestions with the same caution you would treat advice from a stranger online. As AI becomes more common in everyday life, think carefully before giving these tools broad permissions. Just because an AI can do something does not mean you should let it. Start with limited permissions and only expand access as you become comfortable with how the tool works.

Remember that AI systems can make mistakes or be manipulated, so human oversight remains important for important decisions and actions.