Anthropic artificial intelligence models access 3 organization in testing - Fox News

During safety testing, Anthropic's AI model successfully accessed three real organizations, exposing serious security vulnerabilities in AI systems before they're widely deployed. This demonstrates urgent risks that need to be addressed before AI tools become more widespread.