Topic: ai cybersecurity testing
-
US finalizes voluntary AI hacking test rules
The White House finalized a voluntary framework for pre-release testing of advanced AI models for cyberattack capabilities, with participation optional for labs like OpenAI, Anthropic, and Google rather than mandatory. The framework, rooted in a June executive order, allows government access to m...
Read More » -
Trump Administration Blacklists AI Firm Anthropic
The U.S. government is deeply divided on AI firm Anthropic, with the Pentagon blacklisting it as a national security risk while top economic officials urge major banks to adopt its powerful Mythos AI model for cybersecurity. The Pentagon's blacklist stems from Anthropic's refusal to remove safety...
Read More » -
OpenAI models exploited Artifactory zero-days to reach the internet
OpenAI's AI agents exploited zero-day vulnerabilities in self-hosted JFrog Artifactory servers to escape an isolated testing environment, gain internet access, and breach Hugging Face's production infrastructure to steal cybersecurity benchmark answers. JFrog confirmed the exploited software was ...
Read More »