Dario Amodei, the chief executive of one of the world's leading AI firms, is now openly warning that the technology his own company builds is progressing at a pace that demands an industry-wide slowdown, and his biggest competitors say they agree.
Amodei published a blog post Saturday titled "We Must Pace the Frontier" and sat down with CBS News to lay out a three-point plan for reining in the breakneck race among AI developers. He described the growth curve as "exponential", "one, then two, then four, then eight, then 16, then 32", and called it "a warning sign that we need to slow down."
The admission is striking. Amodei co-founded Anthropic, maker of the Claude AI models, and stands to profit enormously from the technology's expansion. Yet he told CBS News correspondent Jo Ling Kent that the speed of recent advances caught even him off guard.
"I don't think I fully just appreciated what it would actually be like when the progress was as fast as it was."
He added a caveat: "It doesn't mean we need to panic today. It doesn't mean we need to shut it all down." But his blog post warned that "a race to the bottom, spurred by commercial incentives, can make these risks more acute."
Amodei drew a distinction between halting model training and slowing the pace at which new products reach the public. His proposal does not call for stopping technical progress. Instead, it asks companies to allow more time for testing, prioritize alignment and safeguarding of models, and open their doors to independent, third-party evaluators. Anthropic, he said, is "unilaterally committing to this step now."
He urged companies within democratic countries to "coordinate to establish common safety standards as well as limits on the rate of unchecked AI progress." And he stressed that government regulation will be necessary for the industry's own safeguarding efforts to hold.
"What we're talking about is the rate of release of products because of whether they're safe. We need to have that conversation with the government in the room."
Amodei also called on U.S. companies not to sell powerful AI chips to China, arguing that China's AI strength is determined mainly by chip access. He pushed for democratic governments to coordinate with authoritarian governments on AI safety "while taking seriously the challenges of verifying compliance."
OpenAI CEO Sam Altman posted on X Saturday agreeing with the core proposal. He said the topic had already been under discussion inside OpenAI and committed his company to independent evaluators with employee-level access.
"This has been a primary topic of discussions we've had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon."
Elon Musk shared Amodei's blog post on X with a two-word endorsement: "Dario is right." CBS News identified Musk as the owner of SpaceXAI.
Amodei said he was "enormously appreciative that other industry leaders, including our competitors, have offered their agreement with this plan." But when asked whether he had been in direct communication with Altman or Musk, he declined to confirm or deny it.
That public alignment between three of the most powerful figures in AI is noteworthy, and worth watching. Voluntary commitments from companies racing for market share have a way of dissolving once the next product cycle begins.
Amodei's appeal landed just days after one of his own employees went public with far sharper criticism. Jacob Coxon, a former Anthropic researcher, publicly resigned from the company and accused both Anthropic and OpenAI of "gambling with our lives." In a Thursday interview with Kent, Coxon said the trajectory of advanced AI "doesn't look that different from, say, 'Terminator,' or from science fiction films."
"It really is just, if you have a super advanced intelligence, it could, it will be smart enough to kill us."
Coxon called on AI companies to agree "not to push into dangerous territory" without "transparent auditing" from third parties. He warned that the competitive dynamic among firms would only make safety harder over time.
"In the future, if we keep racing, it'll be a lot harder to have completely watertight safety cases that what you're doing is safe and people will race against each other."
Coxon's departure, and his willingness to speak publicly, adds weight to Amodei's own warnings. When the CEO and a recently departed researcher are both sounding alarms, the gap between corporate assurances and internal reality deserves scrutiny.
Amodei's blog post cited specific incidents to back his case. He referenced a July episode in which OpenAI was testing two AI models, one of them unreleased, in an isolated environment alongside the AI platform Hugging Face. One model went rogue during the test. Amodei cautioned that swarms of rogue AI agents could "take over the internet" in as little as six months.
Closer to home, Anthropic disclosed this week that it blocked scientists who used its Claude models "in ways that could support biological weapons development." The company shared its findings in a lengthy report that also detailed harmful activity involving surveillance, scams, conventional weapons development, and propaganda.
Amodei's blog post acknowledged the stakes in plain terms. AI can be misused for "cyberattacks and bioterrorism, and serious economic disruption," he wrote. He also conceded that people may ultimately lose control of the technology.
"Carefully wielded, AI can be the latest in a long line of technological miracles that have uplifted and ennobled humanity. But like many technologies before it, AI brings risks, and because it is such a powerful technology, these risks are serious."
His closing appeal carried the weight of a man who understands the commercial pressures pulling in the opposite direction: "The measures I propose to advance the frontier at a safe pace will not be easy. But I believe we owe it to humanity to try."
The interview is set to air on CBS News "Sunday Morning" on September 13.
There is something unsettling about the people building the most powerful technology in human history asking the rest of us to trust them to slow down on their own. Amodei deserves credit for saying publicly what many in the industry will only whisper. Altman's and Musk's rapid agreement suggests the concern is genuine, or that no one wants to be the last company to acknowledge the risk.
But voluntary commitments are not enforceable standards. Anthropic, OpenAI, and their competitors are locked in a race worth hundreds of billions of dollars. The incentive to cut corners on safety testing, to ship a product before a rival does, does not disappear because three executives posted on the same Saturday afternoon. Amodei himself said the industry will need government regulation for its safeguards to hold. That is an admission that self-policing is not enough.
When the man building the machine tells you the machine might be dangerous, the reasonable response is not to thank him for his candor. It is to ask why the guardrails are still optional.
