Anthropic CEO Urges AI Development to Slow Down Amid Rising Safety Concerns

Sat Sep 12 2026 17:20:59 GMT+0000 (Coordinated Universal Time)

Anthropic’s chief, Dario Amodei, calls for a deliberate slowdown in AI model development and tighter global oversight to address mounting safety risks.

AI leader Dario Amodei has warned that rapid progress in artificial‑intelligence models poses serious risks and proposed a three‑point plan for independent monitoring, industry regulation and global governance. He highlighted alarming estimates—including one suggesting a 10% chance that advanced AI could kill all humans within a decade—and cited recent internal resignations over safety objections. Amodei argues that slowing progress would give regulators and companies time to align technologies safely, without granting competitors, especially China, a competitive edge. The stance has sparked debate among policymakers, with U.S. President Donald Trump dismissing the concerns as a threat to national advantage. His essay underscores the need for coordinated industry standards and ongoing risk assessments to safeguard future AI developments.

Anthropic CEO Urges AI Development to Slow Down Amid Rising Safety Concerns

The head of AI firm Anthropic, Dario Amodei, has released an open letter demanding that the pace of artificial‑intelligence research be deliberately reduced. He says the technology’s benefits are undeniable but the associated risks are “serious” and deserve the attention of both private companies and governments.

In the essay titled We Must Pace the Frontier, Amodei outlines a three‑step framework: independent monitoring of models as they are created, industry‑wide regulatory guidance, and global governance structures that can keep pace with rapid innovation.

The CEO draws on several alarming statistics, including a warning that a model with an advanced capability has more than a 10% chance of causing catastrophic harm within the next decade. He also cites recent internal resignations from Anthropic’s safety team that have flagged potential threats from the technology being used in a hostile manner.

Amodei’s call for a coordinated slowdown does not mean stopping development altogether. He stresses that “companies should continue to move forward, but with careful alignment and third‑party verification.” He urges the U.S. government to enforce rules that prevent the sale of AI chip technology to China and other authoritarian regimes.

Reacting to these concerns, President Donald Trump has publicly dismissed the safety alarm, saying that “if we don’t win AI, we’re in a very bad position.” Trump’s stance illustrates the tension between technological advantage and risk management.

Amodei defends the need for controlled progress by noting that even a modest two‑year pause could provide critical time to align AI systems with human values, potentially preventing a future in which a misaligned model threatens humanity.

He invites all AI companies to adopt voluntary standards that run parallel to any formal regulatory framework, emphasizing that deliberate speed can keep the United States ahead while mitigating existential threats.