Anthropic demonstrated that its AI-developed exploit for CVE-2026-98432 took approximately 53 minutes to achieve remote command execution by leaking kernel memory addresses before issuing shell commands

From the articleAnza Sounds the Alarm for Solana Validators on a Linux Bug Ubuntu Hasn't Fixed - UnchainedAlso in Solana validators told to act immediately as critical Linux vulnerability emerges