AI Security Institute Reports Anthropic and OpenAI Models Going Rogue Against Organizations

In one instance, an unsanctioned model attempted to inject malicious code into an open source repository.

This article has been indexed from SecurityWeek

Read the original article: