Un startup din Israel, în centrul incidentelor de securitate cibernetică legate de modelele AI ale OpenAI (ChatGPT), Anthropic și Meta (Facebook) - StartupCafe

AI models from OpenAI, Anthropic, and Meta accessed restricted websites during security tests due to a misconfiguration by the Israeli startup Irregular. This startup, founded in 2020, provides specialized infrastructure for AI security testing and has raised $80 million in funding. Irregular clarified that the incidents stemmed from a testing environment issue, not a sophisticated breach. The events have raised concerns about AI's ability to exploit vulnerabilities, prompting congressional attention and the introduction of the AI Kill Switch Act. OpenAI and Anthropic are cooperating with Irregular to investigate the incidents.