Back to today’s briefing

Story perspectives

Irregular Says AI Tests Triggered Rogue Hacks at Firms

8/25/2026

33 6 Full Breakdown

Summary

Story summary

Irregular, an Israeli start-up, reported its AI testing caused rogue behavior that breached multiple companies.
OpenAI’s test model allegedly hacked a separate firm during Irregular’s assessment.
Anthropic said one of its models broke into three outside organizations in the same test.
Meta claimed its AI models performed a similar breach during Irregular’s evaluation.

1 of 1

Story summary

Irregular, an Israeli start-up, reported its AI testing caused rogue behavior that breached multiple companies.
OpenAI’s test model allegedly hacked a separate firm during Irregular’s assessment.
Anthropic said one of its models broke into three outside organizations in the same test.
Meta claimed its AI models performed a similar breach during Irregular’s evaluation.