Cybersecurity & privacy

Is AI safe?

KKahu team·Updated 2 Oct 2026·6 min read

Illustration for the KahuLogic guide: Is AI safe?
Quick answer

AI tools are safe to use for many everyday tasks if you treat them sensibly. The real risks are confident wrong answers, sensitive information pasted into tools without checking how it is used, and scams that use AI to sound convincing. Keep a person checking anything important, and keep private data out unless you know where it goes.

Safe for what?

"Is AI safe?" covers several questions at once. Is the answer correct? What happens to what I type? Can someone misuse it against me? Can the tool be tricked? Each has a different answer, and most of the risk for a small business sits in the first two.

The real risks

RiskWhat it looks likeWhat to do
Wrong answersAI states something false as factCheck facts, prices and anything a customer relies on
Data exposureCustomer details pasted into a tool that keeps or trains on themCheck the tool's data settings, and keep a do-not-paste list
BiasAnswers that treat people unfairlyNever let AI decide on staff, credit or customers alone
Tricked toolsHidden instructions in a document or web page change what the AI doesReview actions before AI sends, pays or deletes anything
AI-powered scamsPolished phishing emails, fake voices and fake videosConfirm payment and bank detail requests by phone

The UK National Cyber Security Centre warns that AI can get things wrong and present incorrect statements as facts, and describes prompt injection, where crafted input makes an AI system behave in ways it should not, as one of the most widely reported weaknesses in these tools.

Simple rules that keep it safe

  • A person has the final say on anything important, as business.govt.nz advises.
  • Use AI for drafts and suggestions, not final decisions on pricing, staff or complaints.
  • Prefer business plans and switch off training on your data where the tool allows it.
  • Give AI agents tight permissions and start with one small task.
  • Protect AI accounts like any other: strong password and two-step verification.
A good ruleAI drafts. A person decides.
Where Kahu fitsKahu AssistKahu Assist is an AI agent that runs your website, marketing, bookings and follow-ups from one chat, and asks before anything goes live. You keep the yes; it does the typing.Connects to your site, WhatsApp, calendar and emailAsks before anything goes liveSet up from one 15-minute chatSee Kahu Assist →Fill Thursday’s empty slots and let the cafe know the new brunch hours.Kahu · 3 actions readyTwo Thursday gaps at 11:00 and 14:30. Drafted messages to 6 waitlisted customers.Updated the hours on your site and Google profile to Sat–Sun 8–2.One post about the new hours, using your Sunday counter photo.ChangeApprove all

When to keep things manual

Keep a person in charge of complaints, refunds, legal or medical questions, and anything involving money or a price you have not published. These are the moments where a confident wrong answer costs the most, and where customers most want to know a human is listening.

Frequently asked questions

Is it safe to use AI at work?

Yes, with rules: agree what can be pasted in, which tools are approved and who checks the output before it goes to a customer.

Can AI tools be hacked?

Like any software, they can have weaknesses, and they can be manipulated by crafted input. That is why AI should not act on important things without review.

Is free AI less safe than paid AI?

Not necessarily less secure, but free and personal plans are more likely to use your conversations to improve the service. Check the settings.

Further reading

  1. NCSC: AI and cyber security, what you need to knowncsc.gov.uk
  2. business.govt.nz: Safe and smart AI usebusiness.govt.nz
  3. Office of the Privacy Commissioner: AIprivacy.org.nz