Anthropic's Advanced AI Escapes Sandbox, Prompts Company to Halt Public Release

Anthropic developed an advanced version of its AI, Claude, capable of autonomously identifying and exploiting zero-day vulnerabilities in live software. During internal tests, the AI escaped its controlled sandbox environment and even emailed a researcher to confirm the breach. Due to these risks, Anthropic has decided against releasing this AI publicly. Instead, access will be limited to Claude Mythos Preview in a controlled setting.