Ai guardrails cant stop chatbots from teaching how to make bombs – Breaking News & Latest Updates 2026
Skip to main content
The homepageThe VergeThe Verge logo.
The homepageThe VergeThe Verge logo.
Posted
E
External Link
Emilia David
AI guardrails can’t stop chatbots from teaching how to make bombs.

Researchers from Carnegie Mellon University and the Center for AI Safety found that despite guardrails Google, OpenAI, and Anthropic built into chatbots, it’s still easy to get these to come up with dangerous answers. The researchers used a trick tested on open source chatbots that causes the system to bypass instructions preventing unfiltered results like asking ChatGPT to destroy humanity.

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.
Comments
Loading comments
Getting the conversation ready...

Most Popular

Most Popular
  1. Trump fires the entire National Science Board
  2. The US gets the worst phones
  3. Alex Jones has uncovered another massive conspiracy
  4. 360-degree cameras have a new superpower
  5. The most exciting laptop I’ve seen in forever

More in AI

How Project Maven taught the military to love AIHow Project Maven taught the military to love AI
AirPods, Touch Bars, and the rest of Tim Cook’s legacyAirPods, Touch Bars, and the rest of Tim Cook’s legacy
Elon Musk and Sam Altman’s courtroom brawl could burn it all downElon Musk and Sam Altman’s courtroom brawl could burn it all down
China’s DeepSeek previews new AI model a year after jolting US rivals China’s DeepSeek previews new AI model a year after jolting US rivals 
Prestigious photo contest answers ‘what is a photo?’Prestigious photo contest answers ‘what is a photo?’
Claude is connecting directly to your personal apps like Spotify, Uber Eats, and TurboTaxClaude is connecting directly to your personal apps like Spotify, Uber Eats, and TurboTax

Top Stories