What AI means for trust online

The old shortcuts are breaking

For most of history we have relied on quick shortcuts to decide what to trust. A photo meant something happened. A familiar voice on the phone meant a familiar person. A professional-looking website meant a real business. A video call meant you could see who you were talking to.

AI weakens each of these shortcuts. Tools that create realistic images, voices and video, often called deepfakes, are now widely available. This does not mean everything online is fake. It means the cost of faking something convincing has fallen, so we need habits that do not depend on how real something looks.

What the evidence shows in 2026

Official warnings show AI is part of real crime, not just a future worry:

• In September 2026 the FBI warned that criminals pretending to be police and government officials sometimes use fake uniforms, mock offices and AI-enabled video calls to look legitimate. It reported nearly 61,000 complaints about these impersonation scams between January 2025 and July 2026.
• The US Federal Trade Commission warned in September 2026 that criminals are cloning car dealer websites, sometimes using AI, complete with copied photos and fake reviews.
• The FTC has also published guidance for parents on deepfakes and "nudify" apps, which create fake explicit images of real people.

Beyond scams, Google's threat researchers reported in September 2026 that AI is speeding up how fast software flaws are discovered, which helps defenders and attackers alike. The picture is genuinely mixed.

What is being done about it

Labels. Companies are adding ways to show whether content was made with AI, using signed records and invisible watermarks. Google and Apple both announced new features in 2026. Our explainer on AI image labels covers how they work and their limits.

Laws. In the United States, the Take It Down Act requires social media, messaging and photo-sharing platforms to offer a way to request removal of intimate images shared without consent, including deepfakes, and to remove them within 48 hours of a valid request. Other countries are developing their own rules.

Stronger sign-in. Methods such as passkeys cannot be talked out of you by a convincing voice or website, which removes one of the most common prizes for fraudsters.

None of these is a complete answer. Labels are not on everything, laws vary from country to country, and enforcement takes time. That is why personal habits still matter most.

Is it all bad news?

No. The same AI that powers fakes also powers spam filters, fraud detection at banks, and tools that help people with disabilities read, write and communicate. Assistants on phones can summarise long documents and translate in real time. The aim is not to fear AI, but to stop relying on appearances alone.

It also helps to remember that most fraud still depends on old-fashioned pressure: a deadline, a threat, a secret. AI makes the costume more convincing, but the script is the same. Learn the script and the costume matters much less.

Habits that protect you

1. Verify through a second channel. If a relative, boss or bank asks for money or a code, hang up and call back on a number you already have.
2. Agree a family code word. A simple word that only your family knows can quickly confirm a real emergency call.
3. Treat video and voice as clues, not proof. Even a live video call can be faked.
4. Check before you share. Shocking images and clips are the ones most worth checking first.
5. Talk to children and older relatives about deepfakes in a calm, non-judgemental way, so they feel able to ask for help.

Tip: For practical scam spotting, see our current scam alerts. Want the technical version of how social engineering works? Read DoS and social engineering.