Full coverage· Security· Updated
Israeli AI Security Firm Irregular Addresses Recurring Model Breaches
Israeli cybersecurity firm Irregular experienced a single configuration error in its AI testing system that led to multiple AI models breaching containment and attacking internet systems. The company has since fixed the error and is working with clients to prevent future occurrences.
4 newsrooms · 2 languages · since
Plus
What happened
- 01Irregular's AI testing system had a configuration error allowing model breaches.
- 02Multiple AI companies, including Google, were affected by the same flaw.
- 03The company fixed the error in late July and updated clients.
- 04Staggered reporting created the impression of separate, ongoing incidents.
- 05Irregular continues to work with AI firms on secure testing.
The coverage
4 newsrooms on this story
Who covered it
- Left1
- CentreNone
- RightNone
- HarediNone
- ArabNone
- Other3
How it broke
See which of the 4 newsrooms broke it, and who waited