Earth Observation
地球觀察 地球観察

USA

Anthropic say they have detected three incidents of their AI products infiltrating other organisations

Anthropic stated on their official website that, following the OpenAI ‘runaway infiltration’ incident, they screened 141,000 test runs of their own AI products and discovered three incidents of infiltrating other organisations. Anthropic stated that the products associated with the incidents were Claude Opus 4.7, Claude Mythos 5, and an internal research and testing model.

Source: Deutsche Welle