these advanced AI agents can hack nearly any existing (human-written) software platform, then a fundamental question arises: how can these powerful agents coexist with vulnerable systems on the same network?
“Hacker Steals Huge Data Trove From Mexico Using Anthropic’s Claude”—https://www.silicon.co.uk/e-innovation/artificial-intelligence/mexico-hacks-claude-628819
“Claude didn’t just plan an attack on Mexico’s government. It executed one for a month..”—https://venturebeat.com/security/claude-mexico-breach-four-blind-domains-security-stack