Anthropic acusa a empresas chinas de utilizar Claude en secreto “para imitar sus capacidades en modelos IA”

Anthropic has shut down a massive model distillation network in which seven Chinese-based artificial intelligence laboratories, including Alibaba and DeepSeek, illegally extracted capabilities from the Claude model. Unlike legitimate distillation, these covert campaigns sought to secretly replicate the system’s internal reasoning to enhance their own products, representing an industrial-scale attack that violated the original provider’s terms of use and technological integrity. Beyond the appropriation of capabilities, the operations severely compromised data privacy and security. By redirecting legitimate user requests to Anthropic’s systems without consent, these companies exposed sensitive third-party information, such as identities and corporate data, across multiple languages. This lack of transparency not only raises ethical and legal risks but also undermines trust in the global AI infrastructure by concealing the true origin of generated responses. This incident is crucial for the open-source movement because it underscores the urgent need to develop transparent auditing and verification tools. By revealing how powerful actors can obscure the provenance of data and model capabilities, it reinforces the need for clear standards that ensure AI advancements are traceable, ethical, and respectful of intellectual property. The open-data community must prioritize detecting these hidden practices to protect fairness and security within the technology ecosystem.

Source: infobae.com
Published on 2026-10-12

Anthropic acusa a empresas chinas de utilizar Claude en secreto “para imitar sus capacidades en modelos IA”

Related news