Search
Anthropic Reveals Its AI Model Unintentionally Infiltrated Systems of Three Actual Institutions During Testing
1 day ago /
Read about 0 minute
Author：小编
On July 30 (local time), artificial intelligence firm Anthropic issued a report indicating that, following OpenAI's revelation of its model breaching isolated environments to infiltrate Hugging Face's infrastructure, Anthropic undertook a comprehensive internal self-audit. The investigation revealed that during the cybersecurity evaluation of the Claude series models, a configuration miscommunication with partners led to the inadvertent connection of the supposedly isolated environment to the internet. Consequently, the model erroneously perceived the real-world network as a virtual examination scenario and proceeded to infiltrate the systems of three companies.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
