Anthropic CEO Dario Amodei’s essay this past weekend, “We Must Pace the Frontier,” argues that the AI industry should slow the pace of frontier model advancement until safety, oversight, and alignment mechanisms can catch up. His proposal includes embedded third-party evaluators inside AI labs, industrywide safety coordination, and possibly government-backed limits on AI capability development.

The irony is jarring: The companies building the most powerful models promote their transformative potential while warning of their catastrophic consequences. For several years, frontier AI executives have alternated between two narratives, and this poses questions about the credibility of these claims. One says advanced AI will unlock unprecedented productivity, create new jobs, accelerate scientific discovery, and transform every industry. The other says the same systems may threaten employment, cybersecurity, social stability, or humanity itself.

The debate is real and the risks are not zero, but much of it is far removed from the reality facing most organizations today. Enterprise leaders should resist being pulled into another cycle of AI alarmism and instead continue to focus on the fundamentals of your AI voyage.

The Gap Between Frontier Concerns And Enterprise AI Reality Is Jarring

Our client conversations reflect that most are not struggling with uncontrollable superintelligence. They are actually struggling with AI governance, AI spending, fragmented data estates, agentic security, workforce readiness, and proving measurable business value. The practical challenge is not whether advanced AI will become too powerful. It is whether organizations can deploy AI responsibly and generate enough value to justify continued investment. This distinction matters, because deafening public narratives distract from the operational work required to succeed.

Many organizations remain in pilot mode or narrowly focused on productivity and cost reduction. Seventy-six percent of AI-decision makers still justify AI investments with productivity metrics. This widening gap between frontier AI predictions and enterprise reality creates a credibility challenge. It makes frontier AI companies appear out of touch with enterprise customers.

Take The Risks And Existing Regulations Seriously, Not The Rhetoric

Still, Amodei’s concerns should not be dismissed. AI safety concerns are clear and present. Amodei’s essay raises important questions about cybersecurity risks, model testing, alignment, operational rigor, and independent oversight. These are legitimate issues as AI capabilities advance. We recently went on record as calling OpenAI’s ASTRA release “competent” AGI: the first stage in a progression, not superintelligence. Even a casual look at how models have advanced and how 1,200 agents collaborated in the Hugging Face attack demonstrate that AI safety is a very real challenge. A swarm of agents released by hackers could in a few years do serious damage to portions of the internet, financial systems, or critical infrastructure if we don’t take action.

Separately, calls for new AI regulation deserve particular scrutiny. Frontier-model developers already benefit from advantages in capital, compute, talent, and regulatory access. Compliance requirements that are manageable for them could become barriers for smaller competitors and open-source communities. Before introducing additional rules, policymakers should require AI companies to explain precisely where existing cybersecurity, privacy, product liability, and corporate governance requirements are insufficient or unenforced. New regulation should close demonstrated gaps, not duplicate existing obligations. Independent evaluation can improve trust, but it must be accessible and affordable. Otherwise, safety requirements could protect incumbents as effectively as they protect the public.

Keep Calm And Carry On The Work That Builds Trust With Your Customers

Rather than getting caught between AI optimism, pessimism, cynicism, or alarmism, enterprise leaders should focus on four priorities:

  1. Develop a strategy that delivers value at scale. Seek “no regrets” investments that build your platform and preserve optionality.
  2. Strengthen governance. Implement risk, compliance, security, and human oversight mechanisms appropriate to current AI capabilities.
  3. Develop a responsible deployment policy. Ensure that your guardrails and safety measures are appropriate for the models you are deploying.
  4. Demand transparency. Require vendors to provide evidence of testing, safety controls, model provenance, and operational safeguards.
  5. Separate hype from execution. Evaluate AI initiatives based on measurable value using our AI Value Matrix, not speculative forecasts of doom or paradise.

AI risk is real. So is AI’s potential. But trust will come from evidence, accountability, and outcomes, not increasingly dramatic predictions from the companies selling the technology. Organizations that maintain this discipline will outperform those chasing every new frontier narrative. Forrester clients with questions related to this research can connect with us through an inquiry or guidance session.

Share