As far as text goes, steer completely clear of any personal, private, or sensitive information: We've already seen portions of chat histories leaked out due to a bug. As tempting as it might be to get ChatGPT to summarize your company's quarterly financial results or write a letter with your address and bank details in it, this is information that's best left out of these generative AI engines—not least because, as Microsoft admits, some AI prompts are manually reviewed by staff to check for inappropriate behavior. To be fair this is something that the AI developers caution against. "Don’t include confidential or sensitive information in your Bard conversations," warns Google, while OpenAI encourages users "not to share any sensitive content" that could find it's way out to the wider web through the shared links feature.
