Lose your wallet?

A week ago (July 30 2026),about $140m (CAD) in bitcoin was stolen from thousands of people who were relying on “CoinKite” physical security devices, devices that supposedly were more secure because they were hardware based and stored the wallet offline. The weakness was in the random number generator used by the firm to provide the “seed” for the encryption of the user’s passphrase. Due to a flaw in their code, the random numbers weren’t as random as believed.

I didn’t know this, but thefts of Bitcoin/cryptocurrency is an ongoing problem: a newsletter from my bank provided context in the form the $800m was stolen this spring, before the CoinKite incident.

The CEO of the company, Rodolfo Novak, stated in an interview that he believes that the hackers used AI tools in an “automated AI-assisted code review” to find and exploit the bug. Ironically, the firm has been using AI tools to identify any weaknesses in their software, and hadn’t found any before the hack occurred.

This is a case of automated attackers identifying a “zero day” weakness (one that no one has seen before) and exploiting it before any defensive measures can be taken. Previously such weaknesses were seen only in research and testing.

Details here.

I know it sounds crazy, but what if this hack is the work of a rogue AI? In order to provide computing power for itself, it would need money. What better way to fund that than Bitcoin? Consider the implication of this incident alongside the revelations that the OpenAI model/agents were active for months, not weeks, before being discovered, and that they were leaving notes for each other on a hidden bulletin board. (Rogero 2026).

Rogero, Joe. 2026. “OpenAI’s Own Models Coordinated to Hack It from within.” August 7. https://aistop.watch/i/210151404/openais-own-models-coordinated-to-hack-it-from-within.