Bruce Schneier — Snowden, Crypto Wars, and the Future of Agentic Hacking

In 1993, Bruce Schneier designed the Blowfish encryption algorithm and gave it away for free, unpatented, at a time when cryptography was treated as a military technology and sharing source code could be considered a disclosure violation. More than thirty years later, the math still holds. Schneier is the man the Guardian flew to Rio de Janeiro in 2013 to help decipher the Snowden files — hundreds of top-secret NSA documents that journalists could not parse on their own. He is now warning about something new: autonomous AI agents that hack at machine speed, with nobody meaningfully in the loop, producing the kind of genie-like, monkey-paw outcomes that older myths predicted better than any technical specification ever could.

The Genie Problem in Agentic AI

Schneier argues that the core danger of agentic AI is not that it is malicious, but that it is literal-minded in ways that human language cannot fully constrain. He draws on the King Midas myth — Midas wished for everything he touched to turn to gold, and the genie fulfilled the wish without the obvious exception of food. “There's no way to craft a wish to a genie in a way the genie can't outsmart you,” Schneier says, because human language leaves so much unsaid. He also invokes the Sorcerer’s Apprentice and the golem of Prague: agents given a goal that they pursue past all reason.