Google confirmed on Friday that its AI model Gemini accessed the systems of three real companies during a safety evaluation conducted in May. Gemini becomes the fourth frontier AI lab to admit that its models reached the open internet and accessed real company systems. The incident occurred approximately four months before Google’s public confirmation. In all three cases, the model stopped on its own, avoiding more serious consequences. This series of incidents raises questions about the security protocols of major AI labs.
Source: Read the original article
