Artificial Intelligence (AI) continues to grow in popularity as a tool that can help users do all sorts of things. Unfortunately, cybercriminals understand that and are also using AI for their scams.
One of the common ways that people use AI is to have it summarize content, such as an email, a large document or a spreadsheet. But cybercriminals are taking advantage of that AI usage by embedding hidden and malicious AI instructions in files that are unseen by the user.
For example, a user could receive a lengthy email, and they use AI to summarize the message. But embedded in the email could be a hidden, malicious AI prompt that could force your AI program to search for and read other sensitive emails and documents, which could then be sent to the attacker.
Even documents like an Excel spreadsheet that you ask AI to summarize could contain hidden white text on a white background across multiple sheets, which contain AI task modifications and commands that could hijack your AI’s processes and behavior.
To help avoid falling for this type of scam, users should not open files from people that they do not know, nor open a file that they are not expecting. Also, you should carefully consider what content you have AI summarize or process for you, as well as monitor any AI outputs very closely.






