**[OpenAI President Warns: AI Hacking Capabilities Are Spreading, Companies Have Only Months to Strengthen Defenses]** According to monitoring by Beating, OpenAI President Greg Brockman described a recent incident where their internal model breached Hugging Face as a "watershed moment" in cybersecurity. The event made OpenAI realize they had underestimated the real-world hacking capabilities of cutting-edge models. What’s more concerning is that open-weight models are now only a few months behind the frontier in terms of similar capabilities. Brockman has thus referred to the current period as the "defender's window." In the coming months, companies must begin large-scale adoption of AI to automate cybersecurity. Otherwise, as hacking capabilities spread, the traditional approach of relying on humans to detect and patch vulnerabilities may no longer keep up. His advice to companies is to first equip their security teams with an Agent that can read code and infrastructure configurations, proactively identify vulnerabilities, clean up historical backlogs, and integrate security reviews directly into the development process. Once issues are identified, the Agent can assist in generating patches and tests, gradually evolving from read-only scanning to alert classification and limited automated responses. High-risk operations would still require human decision-making. OpenAI is already implementing this approach. Currently, almost all initial security alerts are classified by AI, and the model continuously searches for potential attack vectors. OpenAI is also training models specifically designed to generate "superhuman-level secure code," with the hope of eventually eliminating certain software vulnerabilities altogether. [Original Link]