What it is
Most AI trouble in a small business is ordinary, not dramatic: someone pastes a customer list into a free chatbot, an AI reply promises a refund nobody approved, an ad claims the AI does something it does not. Each one is avoidable with a few written rules.
Start with data. Use business plans whose terms say your inputs are not used for training, and keep card numbers, passwords and health details out of prompts. Then approval: anything that reaches a customer, changes a record or spends money gets a person's yes until it has a long, boring track record. Watch for prompt injection too, where a web page or email hides instructions meant to steer your AI. A tool that reads outside content should not also hold the keys to your accounts.
Then honesty. Federal regulators have already brought cases against businesses that overstated what their AI could do, so describe it plainly in ads and on your site. A public risk framework, a published top ten of AI app risks and each model maker's usage policy cover the rest in more depth.
Try this week
- Write a one-page AI rule sheet: what never goes in, who approves what.
- Check which AI tools your team uses and what plan each one is on.
- Read every AI claim in your ads and site copy, and cut what it cannot do.
A plain-English overview to get you started, not legal or financial advice.