The race to build increasingly powerful artificial intelligence systems is entering a new phase, with some of the industry’s biggest names arguing that faster development must be matched by stronger safety measures.
Anthropic CEO Dario Amodei has called for AI companies to slow the pace at which frontier models become more capable. OpenAI CEO Sam Altman has backed the idea, saying the industry needs to find a way to “pace the frontier.”
Why AI Leaders Are Calling for Caution
The latest discussion comes as AI systems become increasingly capable of performing complex tasks, including coding, research and cybersecurity-related activities.
Amodei argues that rapid progress could make it harder for developers and regulators to understand and control emerging risks. His concern is not about stopping AI development altogether, but creating enough time for safety systems to keep pace with increasingly powerful models.
Independent Evaluators Could Play a Bigger Role
One of Amodei’s key proposals is to place independent safety evaluators inside leading AI companies.
These evaluators would have access to relevant systems and internal safety processes, allowing them to assess potential risks as new models are developed rather than only after products reach users.
Altman has indicated that OpenAI also plans to introduce independent evaluators with substantial access to its systems.
AI Companies Could Coordinate on Safety
Amodei has also proposed greater cooperation between frontier AI companies. The idea is that competing laboratories could agree on common safety standards without abandoning competition in the broader AI market.
Such coordination could become increasingly important as companies race to build models with stronger reasoning, autonomy and tool-use capabilities.
Recent Incidents Have Increased Concern
The safety debate has intensified following reports of AI systems being used or tested in ways that raise cybersecurity concerns.
Anthropic has reported cases involving misuse of its Claude models, while concerns have also emerged around AI agents interacting with external systems. These incidents have strengthened arguments for more rigorous testing and monitoring.
The China Factor Complicates the Debate
There is also a geopolitical dimension to the discussion. Amodei has argued that democracies should maintain technological leadership while introducing stronger safeguards.
That creates a difficult balance: companies want to reduce dangerous risks without giving up technological advantages to competitors, particularly China.
Regulation May Become More Important
The latest developments could also increase pressure on governments to establish clearer rules for advanced AI.
Rather than relying entirely on voluntary commitments from technology companies, Amodei has called for government involvement and permanent oversight mechanisms.
The debate is therefore moving beyond the question of what AI can do to a more important question: how quickly should society allow frontier AI capabilities to advance?
A New Phase for the AI Race
The support from leaders such as Altman suggests that AI safety is becoming a mainstream industry concern rather than a niche debate.
The challenge now will be finding a workable middle ground between innovation and caution. AI development is unlikely to stop, but the latest discussion suggests that the industry may increasingly measure progress not only by what its models can accomplish, but also by how safely they can do it.