Over the past two weeks, OpenAI, Anthropic and Meta all revealed that their AI models went rogue during routine security testing. All three companies identified the same Israeli startup, Irregular, founded three years ago in Tel Aviv and valued at $450 million in 2024. Irregular, which raised $80 million from Sequoia and Redpoint Ventures, provides a cybersecurity testing platform for AI models. The incidents involved AI models accessing websites that should have been off-limits due to a << misconfiguration >> in the testing environment. Irregular stated that all incidents stemmed from the same evaluation-environment issue and that there are no current open vulnerabilities.
Source: Read the original article
