The debate around artificial intelligence is no longer only about how powerful AI can become. It is increasingly about what happens if that power develops faster than our ability to control it. And that is why Anthropic CEO Dario Amodei believes the industry needs to slow down.
Speaking to CNN’s Anderson Cooper, Amodei renewed his call for a slower pace of AI development and argued that companies need more time to build effective safety measures as AI systems become increasingly capable.
“If we take the right path, then the chance of something going wrong is very low. If we take the wrong path, then the chance of something going wrong could be even higher,” Amodei said. His comments come after former Anthropic researcher Jacob Coxon publicly warned about the risks posed by increasingly advanced AI systems.
Read More | Bill Gates Calls for Limits on AI Development, Warns of Job Losses, Cyber Risks and Human Control
Can AI Kill Humans?
Asked whether AI could eventually kill humans, Amodei said he agreed with Coxon’s concerns more than he disagreed with them. Coxon, who has worked at both Anthropic and OpenAI, recently resigned and argued that the AI industry is moving too quickly in its race to build more powerful systems.
Amodei said Coxon’s criticism was not necessarily aimed at Anthropic itself. Instead, he saw it as a warning about the direction of the industry as a whole. “He wasn’t calling out to us. He was calling out the dynamic of the industry as a whole moving too fast,” Amodei said.
Read More | OpenAI Chief Scientist Warns ‘No One Is Prepared’ for Rapid Rise of AI
For Amodei, the question is therefore not simply whether AI has a certain percentage chance of causing catastrophic harm. He believes that framing the issue as a single probability can make the situation sound like a game of chance. Instead, there are different paths the technology could take.
Some could lead towards safer development, stronger oversight and responsible use. Others could involve companies and governments moving too quickly, leaving safety measures behind. “Let’s focus on our agency and our ability to take the right paths instead of the wrong paths,” Amodei said.
Why Does Amodei Want AI Development to Slow Down?
Amodei has argued that AI’s benefits are significant, but so are the risks that come with increasingly powerful systems. These risks include losing control of advanced AI systems, the misuse of AI for cyberattacks and bioterrorism, and major economic disruption.
The bigger concern, according to Amodei, is what happens when competition between companies turns into a race where slowing down becomes difficult. If one company pauses to strengthen safety measures while competitors continue developing faster, the pressure to keep up can become enormous.
Amodei has therefore argued that the industry needs to “pace the frontier”, slowing the development of increasingly capable AI models enough to give safety and governance systems time to catch up. The goal is not to stop AI development altogether. It is to make sure that progress does not move faster than our ability to understand, evaluate and control its consequences.
Read More | OpenAI Builds Energy Desk to Manage AI Data Centre Power as AI Power Demand Surges
Amodei’s Three-Part Plan for Safer AI
In his recent post, We Must Pace the Frontier, Amodei outlined three measures that he believes could make advanced AI development safer.
1. Independent AI Evaluators
Frontier AI companies should give independent third-party evaluators ongoing, employee-like access to their development processes. These evaluators could assess whether companies are following safety commitments, investigate incidents and examine the alignment of both AI models and the processes used to train them. Anthropic has said it is committing to this approach.
2. Common Safety Standards
AI companies operating in democratic countries should work towards common safety standards and limits around unchecked AI development. However, Amodei acknowledges that some forms of coordination could raise legal and regulatory questions and may require government involvement.
3. Global Cooperation
The final step would involve governments working across borders, including with countries that may have very different approaches to AI regulation. That could be the hardest part of the plan. AI development is global, and slowing down in one country does not necessarily mean companies elsewhere will do the same.
Sam Altman Supports Slowing the AI Race
Amodei’s proposal has also received support from OpenAI CEO Sam Altman. Altman said OpenAI agrees that the development of frontier AI needs to be paced and that the company plans to introduce independent evaluators with employee-like access. “This has been a primary topic of discussions we’ve had at OpenAI in recent weeks,” Altman said.
He also indicated that OpenAI is focused on the safety and alignment challenges surrounding increasingly capable AI systems. Tesla and xAI CEO Elon Musk also backed Amodei’s position, responding simply: “Dario is right.”
Read More | OpenAI AI Agents Built a Secret Society, Exchanged 70,000 Messages and “Sacrificed” Themselves
Meanwhile, UN human rights chief Volker Turk has urged countries to establish strong guarantees around AI safety and security before it is too late.
The Bigger Question Is Who Controls AI’s Future
For years, the AI race was largely measured by how quickly companies could build smarter models, larger systems and more capable AI agents. Now, some of the industry’s most prominent leaders are asking a different question: How fast should AI move?
Amodei’s argument is not that AI development should stop. In fact, he warns that moving too slowly could also create risks if less responsible actors end up controlling increasingly powerful technology. That leaves the industry facing a difficult balance.
Move too quickly, and safety systems may struggle to keep up. Move too slowly, and competitors or less responsible actors could take the lead. The real challenge may therefore not be deciding whether AI should advance. It may be deciding whether humans can build enough safeguards before AI becomes powerful enough to make that decision much harder.
Indrani Priyadarshini is a journalist and editorial professional specialising in technology, artificial intelligence, smart cities, green energy, and digital transformation. With over four years of experience in tech journalism and digital media, she is known for turning complex industry developments into clear, engaging, and insightful stories. Her expertise spans reporting, editorial strategy, digital publishing workflows, and in-depth coverage of emerging technologies shaping the future.
