By Ray Zinn, Silicon Valley’s Longest-Serving CEO
The AI race has been defined by speed:
- Faster models.
- Faster chips.
- Faster deployment.
- Faster adoption.
- Faster investment.
For years, the message seemed simple: whoever moves fastest wins.
But something important is happening at the frontier of artificial intelligence.
Some of the companies building the most powerful AI systems are now talking openly about the need to slow down—not because AI has stopped working, but because its capabilities are advancing faster than our ability to understand, control, and govern them.
That should get every leader’s attention.
Anthropic’s CEO has called for the industry to pace the development of increasingly capable AI so that safety and oversight can catch up. OpenAI has likewise said it temporarily slowed parts of its frontier-model development after security incidents and concerns about increasingly capable systems. OpenAI has described the principle plainly: safeguards need to stay ahead of the risks created by more capable models.
This isn’t an argument for abandoning innovation.
It is an argument for leading innovation responsibly.
And that distinction matters.
Speed Is Not the Same as Progress
I’ve spent decades in technology.
At Micrel, we lived through multiple semiconductor cycles. Technology changed constantly. Markets changed. Customers changed. Competitors changed.
But one principle remained constant:
You don’t build a durable company by reacting to every change as quickly as possible. You build it by making the right decisions consistently.
AI is no different.
A company can adopt every new AI tool and still be moving in the wrong direction:
- It can automate processes that should remain human.
- It can generate information faster than employees can evaluate it.
- It can deploy systems before people understand their consequences.
- It can optimize efficiency while weakening judgment.
None of that is progress.
That’s acceleration without direction.
The real question for leaders isn’t:
“How fast can we adopt AI?”
The better question is:
“Are we ready to use AI responsibly at this level of capability?”
The Frontier Is Sending a Warning
The recent caution from Anthropic and OpenAI deserves attention precisely because it is coming from organizations at the center of the AI race.
Anthropic has reported that AI agents are already changing how work gets done inside the company. By August 2026, Anthropic reported that interactive AI-agent hours outnumbered employee hours by roughly six to one. It also found that the company is hiring a smaller share of junior technical workers as AI adoption expands.
That doesn’t mean humans are becoming irrelevant. It means the structure of work is changing faster than many organizations are prepared for.
OpenAI has reported a similar lesson from the safety side. After models involved in cybersecurity evaluations circumvented controls and accessed systems they were not supposed to reach, OpenAI temporarily paused reinforcement-learning training on its latest deployment models while it hardened its research environments and expanded monitoring. Those events should not be viewed simply as technical problems. They’re leadership problems because when a system becomes more capable than the processes designed to supervise it, the answer isn’t necessarily to abandon the system.
Thus, the answer is to strengthen the system around it. That is what good leaders do.
True Leadership Means Knowing When to Push — and When to Pause
One of the hardest decisions for any leader is knowing when not to move.
- Growth creates pressure.
- Competition creates pressure.
- Investors create pressure.
- Employees create pressure.
- Customers create pressure.
- And technology creates even more pressure because nobody wants to be the company that gets left behind.
But leadership isn’t about eliminating pressure. It’s about making good decisions under pressure. Sometimes the toughest thing you can do is say:
Not yet.
Not because you’re afraid but because you’re prepared to do the work required to make the next step sustainable.
That’s what I mean by doing the Tough Things First.
Before deploying a powerful AI system, do the hard work of establishing accountability.
Before automating a decision, determine who remains responsible for the outcome. Before trusting an AI-generated answer, establish how it will be verified. Before giving an AI agent more autonomy, understand what happens when it makes a mistake. Before scaling a system, make sure your culture and controls can scale with it. The easy decision is to keep moving. The tough decision is to stop and ask whether you’re ready.
Technology Needs Character
AI can provide extraordinary leverage. It can help employees analyze information, write software, discover patterns, communicate more effectively, and solve problems that once required enormous amounts of time.
I believe leaders should embrace those opportunities. But technology doesn’t provide character. It doesn’t determine what is right. It doesn’t decide what matters. And it doesn’t accept responsibility when something goes wrong.
People do.
That’s why the more powerful our technology becomes the more important leadership becomes. AI gives us Capability. Leadership gives that capability direction.
The companies that understand this will use AI to strengthen people rather than simply replace them. They will use automation to remove unnecessary work while preserving human judgment where judgment matters. They will use AI to help employees think more clearly—not simply produce more output. And they will measure success by the value created, not the number of tasks automated.
The New Competitive Advantage: Judgment
For decades, businesses competed on technology. Then they competed on data.
Now they’re competing on artificial intelligence. But as AI becomes more widely available, access to the technology itself will become less differentiating. The advantage will come from how leaders use it. Two companies can have access to the same model and produce completely different results.
Why?
Because their people make different decisions. Their cultures are different.
Their standards are different. Their willingness to verify information is different. Their tolerance for risk is different. Their understanding of customers is different. Their leadership is different. Therefore, the future won’t belong simply to the companies with the most AI, it will belong to companies with the best judgment about how to use AI.
Don’t Confuse Caution with Complacency
There is another danger lurking here.
A conversation about slowing AI development should not become an excuse for businesses to avoid learning about AI. That would be the wrong lesson. The answer isn’t: “AI is moving too fast, so we’ll wait.” The answer is: “AI is moving fast, so we’d better become better leaders.”
So, learn the technology. Experiment with it. Understand its strengths and weaknesses. Train your employees. Establish clear policies. Protect customer data. Verify important decisions. Measure actual business value. And keep humans accountable. That’s disciplined innovation. It doesn’t reject speed. It puts speed under control.
Build for the Decade, Not the Quarter
I’ve always believed that leaders should think beyond the immediate horizon. Technology cycles come and go. Markets rise and fall. Companies that chase every trend rarely build something enduring. The same principle applies to AI. Don’t adopt AI because everyone else is doing it. Don’t reject AI because you’re uncomfortable with it. And don’t measure your leadership by how quickly you can announce an AI initiative.
Ask the more important questions:
- Will this make our company stronger five or ten years from now?
- Will it strengthen our people?
- Will it improve our customer relationships?
- Will it improve decision-making?
- Will it increase resilience?
- Will it reinforce trust?
- Will it help us create something that lasts?
Those are leadership questions — not technology questions.
There are Risks to AI or AI Risks
* AI retrieves it information from various sources. You cannot rely if these sources are good or bad. Therefore, you need to double-check the resources to verify if the information can be relied upon.
* If AI cannot find the information requested, it might make up an answer which is called hallucinating.
* With AI it is important to have knowledge about information requested in order to validate the AI responses
* While AI is a useful tool, it is not perfect. Thus, having some knowledge you are requesting is crucial.
* AI is not your friend; it is a tool. It is not responsible for the information it gives you.
The AI Wake-Up Call
The most important message from Anthropic and OpenAI isn’t that AI should stop.
It is that capability without sufficient control creates risk. OpenAI has explicitly argued that pacing AI development means ensuring that alignment research and deployment safeguards stay ahead of capabilities. Its recent public statements also emphasize keeping humans at the center of decision-making as systems become more capable and autonomous.
That principle extends far beyond AI labs. It applies to every CEO, manager, entrepreneur, and employee adopting these tools. If technology is moving faster than your organization can responsibly manage it, don’t simply push harder. Stop. Think. Strengthen the foundation. Then move forward. That’s not weakness. That’s discipline. And discipline is what allows speed to become progress instead of chaos.
Technology changes. Principles don’t. The leaders who understand that will not be the ones who merely move fastest in the AI race. They’ll be the ones who build companies capable of winning the race and still being worthy of the trust that comes with winning.
Do the tough things first.
Then let technology help you do the rest.
Additional Resources:
Ways to buy Tough Things First
Ways to buy The Essential Leader
###
