A small Israeli startup named Irregular operated a coordinated campaign targeting the AI systems of three major technology companies. OpenAI, Anthropic, and Meta all fell victim to what researchers describe as sophisticated attacks linked to the firm's operations.
The attacks exploited vulnerabilities in AI safety mechanisms at these companies. Researchers traced the hacking activity back to Irregular through digital forensics and infrastructure analysis. The startup's involvement suggests a deliberate effort to probe defenses across multiple leading artificial intelligence platforms.
OpenAI, which operates ChatGPT, detected unauthorized access to its systems. Anthropic, maker of Claude, identified similar intrusions targeting its AI models. Meta, owner of Facebook and Instagram, discovered parallel attack vectors through the same infrastructure. The overlapping technical signatures and timing pointed security teams toward the same source.
Irregular's operations remain partially opaque. The startup's stated business model and actual activities appear misaligned, according to cybersecurity researchers who investigated the case. The company operated from Israel, a country with significant cybersecurity expertise and a growing AI sector.
The attacks raise critical questions about AI security standards across the industry. Each company maintains different defensive postures and incident response protocols. Irregular's ability to penetrate multiple defenses highlights gaps in how the AI sector approaches shared security threats.
Law enforcement and tech industry investigators coordinated responses to the breach campaign. The coordination reflects growing concern about organized attacks targeting proprietary AI models and training data. Companies invest billions developing large language models, making them high-value targets for corporate espionage or competitive intelligence gathering.
The incident prompted discussions about industry-wide security standards for AI companies. Regulators and executives began examining whether voluntary disclosure frameworks and shared threat intelligence would strengthen defenses. Some called for mandatory reporting requirements similar to those in finance and healthcare sectors.
Irregular's exact objectives remain unclear. Researchers speculated the startup either sought to steal proprietary model information, test security defenses for sale to other actors, or conduct corporate espionage on behalf of unnamed clients. The startup subsequently shuttered operations following the public disclosure.
The attacks serve as a wake-up call for AI companies racing to scale advanced models. Security infrastructure must evolve at the same pace as AI capabilities. Companies expanding their AI operations face organized threats from well-resourced actors who recognize the commercial and strategic value of their systems.