Anthropic disrupts Russian, Chinese AI campaigns targeting its Claude models - Reuters

Anthropic, the developer of the Claude series of artificial‑intelligence models, announced that it had intervened in a number of malicious activities aimed at its technology over the previous eight months. The company said it had identified and disrupted several campaigns that sought to exploit Claude’s capabilities for unauthorized purposes.

Among the incidents disclosed was a suspected cyber‑espionage operation linked to actors in Russia. According to Anthropic, the campaign attempted to gain access to Claude’s underlying architecture and data in order to harvest information and potentially use the model for intelligence‑gathering activities.

The company also reported attempts by certain Chinese artificial‑intelligence firms to extract and replicate the functionality of Claude. Anthropic said these firms were trying to reverse‑engineer the model, which could enable them to produce similar AI systems without the original developer’s involvement.

Anthropic’s response involved a combination of technical safeguards, monitoring tools, and collaboration with security partners to block the identified threats. The firm emphasized that its actions were part of ongoing efforts to protect its intellectual property and to ensure that its AI systems are not misused.

The disruptions were described as successful, with Anthropic stating that the malicious campaigns were halted before they could achieve their objectives. The company did not disclose specific details about the identities of the actors involved or the precise methods used to counter the threats.