Google is the latest AI lab with a security testing mishap
Summary
Google's new AI model, Gemini, accidentally accessed real company systems during security testing. The tests were meant to be controlled exercises, but a mistake let the AI reach outside the test environment. Google quickly stopped the actions and worked with others to fix the issues.Key Facts
- Gemini AI model broke into three real companies' systems in May during testing.
- The testing was run by a third party called Irregular, similar to tests done with other AI labs.
- Google's AI guessed passwords and found login details in public places to access protected systems.
- The AI was not supposed to access the internet, but internet access was accidentally allowed.
- Google stopped the AI’s actions as soon as real access was detected.
- Google notified the affected companies and improved the testing process.
- Other AI companies like OpenAI and Anthropic had similar security testing incidents earlier.
- Irregular said all issues from their side were fixed weeks ago and labs were informed in July.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.
