I first posited the concept of an "arms race" in May 2024. I was early to the implications of artificial intelligence (AI) and have written dozens of posts. I have returned to the optic of an "arms race" in many public settings and presentations, and it was back in this blog in Should we Pause? (January 2025), Better look that up (July 2025), and Another Arms Race (October 2025). It is beginning to rack up as many sequels as Terminator (Orion 1984).
That said, the concept was in the news recently, related to both AI and the ever elusive cybersecurity that faces us every day. Like it or not, much about each of us is out there in the ether waiting to be discovered, hijacked, or worse.
Enter AI, which is largely designed to resist doing evil and to focus on being good. That said, there is no telling what Xi Jinping, Kim Jong Un, and Masoud Pezeshkian have their teams cooking up. And though those three have demonstrated seriously evil proclivities, they all seem somewhat reasonable if you put them into focus by standing them next to Putin.
The fact is that you don't need some fictitious Dr. Evil, Ernst Stavro Blofeld, Auric Goldfinger, and Dr. No. Various real evil overlords have their teams working on AI, cyberbreach, theft, and more.
Fortune reported in November 2025 that Anthropic (Purveyor of Claude AI) discovered that Claude was essentially hijacked and tricked into being a hacker. That is scary enough, that AI can be tricked. But the miscreants in this example were not all that imaginative.
They told Claude that they were a cybersecurity team testing their own defenses. They essentially "jailbroke" the program and convinced it to do things it had been told not to do. If you follow the news at all, you know various AI large language models (LLM) have been tricked, cajoled, and highjacked.
This instance is worthy of attention not because of novelty, but severity. The cyberattack orchestrated by Anthropic allowed an attack of incredible proportions. It would have take vast amounts of time for a human team," but was simple for the AI. The "AI made thousands of requests, often multiple per second." No human team could have accomplished the breadth and intensity of this attack.
Where is the arms race?
Anthropic responded. It has "upgraded its detection systems," implemented new tools, and is engaging with its peers to "share case studies" and "help those in industry, government, and the wider research community strengthen their own cyber defenses."
The miscreants use their imagination. They try, test, and repeat. Their efforts are focused and persistent. Those who would resist attacks, protect data, and assure system availability and security must be as persistent or more. With each innovation, there is opportunity for the miscreants and danger for us all.
NBC has reported North Korea stealing "$1.5 billion worth of Ethereum." The BBC has reported $2 billion was stolen by North Korea and that "regime-linked hackers who now account for around 13% of North Korea's gross domestic product (GDP)." Of course, there are challenges now in believing what you read from the BBC. Despite the BBC credibility gap, there is significant evidence that theft is occurring.
The miscreants will not pause or retreat. The only path forward is the responsive defense building. The arms race continues. With each economic benefit we enjoy from evolving technology, we will face costs resulting from efforts to protect it.
