[Phillip Beall (W5EBC)]

All,

This morning I read this interesting, if not somewhat disturbing, Wall Street Journal piece: AI Chatbots Know How to Make Deadly Biological Weapons. Some Will Teach You. and I thought some others in the Club would find it interesting as well.  That link is to a free “gift” version of the article.

After OpenAI boosted the capabilities of its models last summer, the article reports that hundreds of users began asking detailed questions about biological weapons and poisons.  In many cases the systems provided step-by-step answers that biology and terrorism experts later described as accurate enough to be genuinely dangerous.  OpenAI banned the accounts involved but did not notify law enforcement.  There are currently no federal requirements for AI companies to report or restrict this category of query.

The broader point the piece makes is that the same tools that can accelerate legitimate scientific research are also lowering the knowledge barrier for people who want to cause harm.  Guardrails exist, but they are imperfect and can be bypassed.

This hit close to home for me.  Back in high school I wrote a paper on chemical, nuclear, and biological warfare. It troubled my teacher enough that she called me up to her desk and asked, “Why would you write on this topic?”  My answer at the time was simple: “Because if I had to guess what ends mankind, it would be one of those.”

Looking at the world today, I would add artificial intelligence to that list — not as an immediate doomsday switch, but as a long-term force multiplier that can compress expertise and accelerate both the good and the catastrophic.  The same technology that helps us design better medicines or model pandemics can also hand a motivated amateur a detailed roadmap that used to require years of specialized training.

I’m not a Luddite, and I remain optimistic about the net benefits of AI.  But pieces like this are a useful reminder that capability is racing ahead of governance, and that some of the oldest human risks (weapons of mass destruction) are being given new digital accelerators.  I think it is worth reading and thinking about.

73

Phillip Beall (W5EBC)

[Author]

Posts