A Single Firm is Behind OpenAI, Anthropic, and Meta Hacking Scandals

An investigation by Brian Chau at Effort News suggests that recent "hacking scandals" involving OpenAI, Anthropic, and Meta, initially framed as "rogue" frontier models, were actually due to basic, sloppy DevOps and botched testing by a single Israeli third-party vendor. This report rebrands what was perceived as an apocalyptic near-miss into a case of ordinary IT incompetence, implying that such incidents are sometimes weaponized to justify regulatory capture rather than indicating an imminent Skynet scenario.

This report is the first to link a single firm, Irregular, to hacking incidents across OpenAI, Anthropic, and Meta, unlike previous disclosures that focused on individual company breaches.