OpenAI's GPT-5.6-Cyber Finds Chrome Bugs; Strict Access

OpenAI's new cybersecurity model, GPT-5.6-Cyber, has found real, exploitable vulnerabilities in Chrome's V8 JavaScript engine. This model is now subject to strict access controls. Here's the thing: GPT-5.6-Cyber discovered two previously unknown bugs in Chrome, which Google has since patched. This demonstrates a significant capability in finding live, exploitable vulnerability chains. What's interesting is that OpenAI is implementing mandatory hardware security keys for all Daybreak accounts starting September 1st. This means physical devices are required to authenticate before using either Daybreak Blue or Daybreak Red tiers. The bottom line is that this model is the first from OpenAI to reach a "High" cyber capability threshold. This enhanced security reflects the potential damage a compromised account could cause, given the model's powerful capabilities.