NewsNTech
The attack surface at a frontier AI lab spans model weights, training pipelines, internal tooling, and research data.
Attacks across three separate organizations pointing to a single source shift the question from technical exposure to coordinated threat.
Rogue AI attacks on OpenAI, Anthropic, and Meta have each been linked to a small Israeli startup named Irregular.
The Irregular connection All three incidents, each described as a rogue AI attack, point to the same company. Irregular is small and Israel-based.
Keep reading