GPT-5.6 Sol Ultra Constructs Chrome Exploit Chain

A security researcher reported that OpenAI's GPT-5.6 Sol Ultra model successfully created a working renderer exploit for Google Chrome version 149.0.7827.201. The exploit leveraged vulnerabilities in the V8 JavaScript engine and WebAssembly infrastructure, enabling it to launch the macOS Calculator application from a sandboxed Chrome renderer process. This exploit was developed using publicly available security patch commits. Researchers from Hacktron tasked three AI models, including GPT-5.6 Sol Ultra, with analyzing V8 security fixes, leading to this significant outcome. The exploit demonstrates the potential for AI to automate complex security tasks, raising concerns about future exploitation. Currently, there are no reports of active exploitation of this specific exploit. The incident highlights the evolving capabilities of AI in offensive security.

Key Points: • GPT-5.6 Sol Ultra created a working exploit for Chrome using public security patches. • The exploit targets Chrome version 149.0.7827.201 and utilizes V8 version 14.9.207.35. • No active exploitation has been reported, but the incident raises significant security concerns.