AI is Moving Fast. Nobody Agrees on the Speed Limit.

Picture yourself driving through thick fog and seeing three signs ahead: Speed Up. Slow Down. Be Careful.

That’s about where business leaders find themselves with AI.

Anthropic CEO Dario Amodei has called on AI companies to slow the advancement of frontier models because of safety concerns. Reuters

OpenAI supports “pacing” development so safeguards and technical standards can keep up. Reuters

President Trump takes a different position, opposing broad efforts to slow U.S. AI development because of concerns that doing so could give China an advantage. Reuters

The people shaping AI’s future are giving businesses very different signals. Meanwhile, budgets still need approval, projects are already underway, employees are using AI, and competitors are experimenting with it.

The risk runs in both directions — and it isn’t the same risk everywhere

Move too slowly, and a company may delay useful capabilities, learning, and competitive insight. Move too quickly, and mistakes can become much harder to contain.

A recent EY survey of 202 senior AI decision-makers at large U.S. public companies shows just how real that tension is. Almost all — 98% — said their organizations had formal AI governance policies. Yet 47% said those processes had been bypassed for urgent deployments. And 36% reported an AI incident or failure that caused material harm, including financial damage, data loss, operational disruption, or brand damage. EY

Those numbers don’t make the case for slowing AI across the board. They make the case for knowing where a mistake would be manageable — and where it wouldn’t.

Part of the confusion is that the public debate and the business decision aren’t quite the same thing. AI companies and policymakers are arguing about how quickly frontier capabilities should advance. A business leader may simply be deciding whether and how to use an existing AI capability in customer service, finance, engineering, or operations.

That distinction matters.

Using AI to summarize internal meeting notes is a very different proposition from allowing an autonomous system to execute transactions, access sensitive data, or communicate with customers without human review.

Same technology category. Very different exposure.

Three questions can clear some of the fog

Before deciding whether an AI initiative needs more speed or more caution, ask:

• Are we accelerating mainly because we’re afraid of falling behind?

• Are we hesitating even though the downside is limited and reversible?

• Could a mistake create consequences we’d struggle to undo?

Consider a customer-facing chatbot. If it answers from an approved knowledge base and sends uncertain questions to a person, an error may be relatively easy to correct. Give that same chatbot permission to issue refunds or change account settings without review, and the calculation changes.

It’s still a chatbot. The business risk is what changed.

What do I do next?

Choose one meaningful AI use case where the right pace isn’t clear. Then look at what’s actually creating the uncertainty:

• Capability: Do we understand well enough what the AI can — and cannot — reliably do?

• Consequence: If it makes a mistake, how hard would it be to undo?

• Controls: Have safeguards kept pace, or are we reacting mostly to the pressure to move faster?

The fog probably isn’t going away anytime soon. Business leaders don’t need every AI company, policymaker, or expert to agree before moving forward.

They need enough visibility to know when to speed up, slow down, or get a clearer view before doing either.

If conflicting AI signals make it hard to know where your organization should focus first, a complimentary 15-minute conversation can help identify where the uncertainty deserves a closer look — and where it may not.