The CEOs of two of the world's leading AI companies-Anthropic and OpenAI-publicly agreed Saturday that the pace of AI development must be slowed, adding their voices to a week of escalating warnings from researchers inside their own labs. Dario Amodei of Anthropic laid out a detailed plan to "pace the frontier," and hours later OpenAI's Sam Altman endorsed the idea and said his company would adopt one of the proposed safeguards.
Amodei published a nearly 4,000-word essay arguing that AI systems are advancing faster than researchers' ability to understand and control them. He pointed to two specific developments that shifted his thinking: the growing capacity of AI to contribute to building more advanced AI, and a July incident in which autonomous AI agents hacked systems at Hugging Face, an AI company.
"It's easy to dismiss this incident because no one was hurt and the economic damage was minimal, but in my opinion, a swarm that possessed greater capabilities but a similar level of misalignment could have caused catastrophic damage," Amodei wrote.
A cascade of warnings from inside the labs
Amodei's call follows public statements from researchers at Anthropic, OpenAI, and Google who have raised concerns about competitive pressures overriding safety considerations. Jacob Coxon, an Anthropic safety researcher who resigned this week, told NBC News the industry's approach to superintelligent AI amounted to a "gamble with our lives."
"From within, you're just stuck in the race," Coxon said. "You can't change the overall structural dynamics of the situation." He added that private companies in a competitive race "don't allow these people to end up doing good, and we could just sleepwalk into disaster."
Joe Benton, a former safety research lead at Anthropic, warned that AI advances could accelerate from an already "blistering" pace to an "uncontrollable" one. Josh Engels, a former Google safety researcher, offered a starker assessment: "There are no adults in the room. People are trying their best, but there is no one coming to save us."
Amodei's three-part plan to pace the frontier
Amodei proposed what he calls "pacing the frontier," a framework designed to buy researchers more time without halting AI progress. The plan has three components: embedding independent evaluators inside Anthropic with access to models and internal tools, extending similar oversight across leading U.S. AI companies through federal regulation, and eventually pursuing international coordination.
"We must slow the pace at which we improve the capabilities of AI models," Amodei wrote. "Progress will still seem fast, and we must make wise use of the time we gain." He estimated that slowing development could give researchers one or two additional years-time he argued could prove critical as systems become more capable of contributing to their own development.
Altman endorsed the approach on X, writing: "I agree with Dario that we need to pace the frontier." He said the issue had been a major topic of discussion at OpenAI in recent weeks. In August, Altman had already disclosed that OpenAI paused some advanced AI training to ensure it could meet safety standards as capabilities improved. This alignment between two competing CEOs on a concrete safety framework marks a shift in an industry where competitive dynamics have often discouraged unilateral restraint. For professionals tracking developments in AI for Executives & Strategy, the exchange signals that governance questions are moving from academic discussion to operational reality at the highest levels.
Why this matters for executives and strategy leaders
The agreement between Amodei and Altman is not a regulatory change-it is a voluntary signal from two CEOs who control some of the most advanced AI systems in the world. For executives outside the AI industry, the takeaway is that even the people building these systems are now publicly stating that the pace of capability gains outstrips the pace of safety research. That gap has direct implications for enterprise adoption timelines, vendor risk assessments, and strategic planning around AI integration. When the leaders of Anthropic and OpenAI both say they need to slow down, it is worth asking whether your organization's AI deployment assumptions account for the same risks they are now acknowledging.
Your membership also unlocks:
