AI Fake News is Getting Harder to Spot — Here’s the Risk
Your brain is losing the fight against AI-written lies. Good news: there’s a way out.
Deep Dive
A new study with 504 participants and 2,438 judgments found that people often can't tell AI-generated news from human-written news — and heightened suspicion doesn't improve detection. Under sustained exposure, fake-news detection accuracy dropped by 10.2 percentage points, while spotting AI origin stayed stable. The researchers mapped these results using an adapted cybersecurity kill chain to identify candidate intervention points for proactive defense against AI-driven disinformation.
Key Points
- AI can now write fake news that fools even careful readers most of the time
- Fake-news detection drops by 10.2% after prolonged reading, but AI-origin detection stays flat
- Researchers propose using cybersecurity-style “kill chains” to block disinformation early
Why It Matters
AI-written lies are getting harder to spot and spreading faster than ever, putting your trust — and democracy — at risk.