AI firms link recent test anomalies to misconfigured environments hosted by Tel Aviv-based Irregular, raising cybersecurity concerns.
OpenAI, Meta, and Anthropic identified security test anomalies in their AI models, attributing the issues to test environments hosted by Israeli startup Irregular. The misconfigurations exposed models to unintended web targets, increasing cyber risk exposure for the firms.
The incidents highlight vulnerabilities in third-party AI testbeds, with prior reports of similar breaches remaining limited. High-profile AI companies have faced growing scrutiny over security protocols, prompting legislative proposals like the AI Kill Switch Act to mandate emergency shutdown capabilities.
While the firms have not disclosed financial impacts, the disclosures may heighten regulatory and investor focus on AI infrastructure security.