GPT-5.6 Sol Ultra Builds Full Chrome Exploit Chain From Security Patches

A frontier AI model (Sol Ultra) analyzed V8 security-fix commits and independently developed a full Chrome exploit chain—building primitives (addrof/fakeobj), expanding sandbox read/write, leaking native addresses, abusing a Wasm NativeModule use-after-free to gain native primitives, pivoting the WebAssembly code pointer table, and achieving code execution (launching Calculator). The report highlights that this demonstration shows how large models could scale exploit development from patches, increasing the urgency for rapid patching and improved defensive detection.