Guide · Beginner

Five Everyday AI Pitfalls and How to Protect Yourself

GBA AI Safety & Development Standing Committee · 2026

AI has merged into daily life — customer service, translation, writing, cameras, banking, health apps. Behind the convenience sit a few common pitfalls everyone should know. This article builds your first line of defence in the plainest possible terms.

Pitfall 1: fake customer service and phishing links

Scammers impersonate banks, couriers, government offices or AI-platform support, sending links by SMS or email — "account anomaly", "parcel held", "refund — click here". One click and a phishing site harvests your login password or card details.

Protect yourself: never log in via links in SMS/email. Open the official app or type the official address yourself; real support staff never ask for your password by text.

Pitfall 2: over-permissioned apps

On install, many apps demand contacts, location, photo library and microphone access — mostly unused. One tap on "Allow" and your data may quietly leave.

Protect yourself: grant only necessary permissions; switch the rest off in system settings. Review "permission manager" regularly and revoke what you don't need.

Pitfall 3: reusing the same password

One password across dozens of sites is the most common serious risk — one breach and criminals "lock-pick" your other accounts with the same credential.

Protect yourself: one account, one strong passphrase. Use an open-source password manager (e.g. KeePass, Bitwarden) to store and generate them, and switch on two-factor authentication (2FA).

Pitfall 4: deepfakes — fake voices and fake images

AI can now generate uncannily real faces and voices. Fraudsters may clone "your friend's voice" to phone you for a transfer, or fabricate news images to spread rumours.

Protect yourself: verify important decisions (especially transfers and identity checks) through a second channel — call back yourself, confirm face-to-face. See a suspicious clip or image? Doubt first, forward later.

Pitfall 5: blindly trusting AI answers (hallucination)

Large language models sometimes "lie with a straight face" — inventing facts, figures and links that do not exist. This is "hallucination", and it is especially dangerous for health, legal and financial information.

Protect yourself: treat AI as a "clever first draft", not an authority. Always cross-check important information against official, trusted sources — never rely on AI's word alone.

One-line summary: the three locks

Use AI safely by remembering three principles — Safety (recognise risk), Reliability (verify sources), Trustworthiness (control your own data). Build these habits and you already protect yourself better than most.

Want to know more? Ask our AI Safety Assistant free of charge, or send your questions to info@3token.org.