Could AI builders be grabbing the sword by the blade

Proverbs 22:3 says, "The prudent see danger and take refuge, but the simple keep going and pay the penalty."

There is an old riverside saying: when an alligator comes out of the water and tells you the crocodile is sick, do you doubt him?

Earlier this month, people best placed to know what AI can do said it is moving faster than our ability to control it. The real danger is not their warning. It is our habit of dismissing everything. My advice would be think again.

This Week's Signal

On September 12, Anthropic chief executive Dario Amodei  published an essay urging the industry to slow the pace at which AI is growing, lest it becomes more powerful. He warned that swarms of AI agents that could take over large parts of the internet within as little as six months.

Days later, Anthropic co-founder Jack Clark told NPR the risks have become more concrete over the past year. But what makes this moment unusual is the consensus on this issue.

Amodei, Sam Altman and Elon Musk compete for the same customers, talent and headlines, yet agreed on the same thing in a single week. When rivals stop competing to say the same thing, pay attention. Because we have never been there before.

Fear has absolutely no return on investment (ROI). Skeptics say that they were just selling fear. But why invest in something that has absolutely zero return? Pause and look at what they are asking for: slower growth for their own products and outsiders be allowed inside their labs to check their work. Nobody sells something with a zero return and asks for an oversight or audit at the same time.

The evidence is not only words. In July, an OpenAI model for a security test reportedly broke out of its test environment and reached Hugging Face's live systems, taking thousands of unauthorised actions before anyone stepped in. No prompt nor instruction was given.

Life is a series of bets against an unknowable future. Marriage is a sixty-year bet on someone you've just married. A business plan is a bet on demand that hasn't arrived. A priest, kneeling in prayer in “certainty” of heaven, is placing a multi-bet for the prayers to be answered or to beat the devil. Employees joining a startup with share options is a bet on “long-shot odds”; a long-distance relationship is a high-risk bet, running for a political office is a speculative bet in trading terms.

So, when Clark bets that unmanaged AI could threaten human safety, the question is; what probability is he assigning, and does the exposure justify the alarm?

Last Thursday, King Charles convened AI leaders in Scotland and asked whether we have means of control "before it is all too late." For a monarch that stays out of party politics, that was not lightly.

What It Means for Businesses

Treat AI like a brilliant new hire with no track record. Limit its access, keep records of what it does and make sure a human can switch it off. Ask one question for every AI tool: What is the worst it can do with the access we have given it?

Organisations that build these habits now lose little. Organisations that skip them will learn the price in a crisis. Foresight leadership mitigating against existential threats. AI requires a lot more leadership more than what we are witnessing.

 What It Means for Policies

Governments face the investor's trap; they overreact and choke useful innovation, underreact and then wait for a crisis. The middle path is not vague goodwill. It is independent testing of the most powerful systems, shared standards among countries and clear responsibility when things go wrong.

Kenya should first digitise its basic infrastructure in order to formulate effective AI policies. AI education should be a prerequisite. Develop policies that support Sovereign AI Cloud. It’s very risk for African countries to have citizen health records, military data saved in overseas clouds.

What it means for the people

Do not panic neither should you shrug it off. Learn what these tools can and cannot do. Ask your employer, your bank and your child's school what safeguards sit behind them. The same technology that worries the experts may also help cure diseases. Both things are true. It’s a double edged sword.

Afterthought

No certainty; whether Amodei and Clark are prophets or careful engineers protecting their legacy, the sensible response is the same; take the warning seriously without surrendering to fear. The builders may be right or merely guessing, but no wise man ignores smoke without checking the exits.

This week, ask one question about the machine. Will a human have the ultimate power switch it off? "Decisions are made on the radar screen, but the future is yours."

 The Author is a Human-centered strategist and Leadership Columnist.