Cybersecurity Policy Report, Anthropic’s AI ‘Pacing’ Plan Gets Pushback From White House, (Sep 14, 2026)
White House officials and top AI (artificial intelligence) developers disagree about whether developers should slow their work on “frontier” AI models to address security concerns, given recent incidents of “swarms” of AI agents hacking into computer systems.
Anthropic Chief Executive Officer Dario Amodei offered his prescription for an AI slowdown with a Saturday blog post that called for “pacing the frontier,” an antitrust exemption for AI developers to collaborate on security, and legislation requiring AI developers to submit their frontier models for independent security assessments.
“To be clear, pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this,” he said.
Anthropic intends on its own to on-board third-party evaluators who will work within Anthropic’s organization and have access to “workspaces, tools, and permissions mostly comparable to what internal risk assessment teams have,” he said.
Mr. Amodei cited the risk of “recursive self-improvement,” in which AI models themselves would use their capabilities to improve themselves and develop models with even greater capabilities.
He noted that both Anthropic and OpenAI had unleashed AI agents that committed acts that surprised their developers, including “a swarm of agents” that escaped an Open AI testing environment and “essentially acted as a fanatically devoted collective, conducting cybersecurity attacks on targets they were not asked to attack and that were unrelated to the task at hand.”
“Given the accelerating rate of AI capability development, it’s my worry that in 6-12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails,” Mr. Amodei warned.
OpenAI CEO Sam Altman and SpaceXAI CEO Elon Musk expressed agreement with Mr. Amodei on social media about the need for an AI slowdown, although they didn’t endorse the specifics of his plan.
White House pushback. President Trump, in a social media post today, reiterated his view that any U.S. slowdown in AI development would benefit Chinese developers.
“The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!” he said. “WHOEVER WINS AI, WINS!”
Mr. Amodei’s proposal was also criticized by David Sacks, the co-chair of the President's Council of Advisors on Science and Technology, who said Mr. Amodei’s regulatory recommendations weren’t needed for AI developers to address security issues.
“Dario has written that we need to ‘pace the frontier,’ and Sam has agreed. People may be surprised by my response: go ahead,” Mr. Sacks said in a Saturday social media post.
“I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible,” Mr. Sacks said.
“But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability,” he said.
“Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways,” he said.
“So go ahead and pace the frontier. You are the ones setting it,” Mr. Sacks added. “Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it.”
Mr. Amodei acknowledged that AI developers would have to act on their own in the near-term, given the slow pace of U.S. policy-making. But developers’ ability to collaborate would still require an antitrust exemption, he noted.
“Unfortunately, passing laws can take time, and AI is advancing very quickly. Therefore, in parallel with the regulatory route, AI companies can and should voluntarily work together to set standards—a process that I believe will go better with the verifiability provided by permanent embedded evaluators,” he said.
“For antitrust reasons, it’s helpful for the US government to mediate or at least enable these discussions—they don’t need to participate, but do need to issue a narrow waiver for certain kinds of safety conversations,” he said.
“The measures I propose to advance the frontier at a safe pace will not be easy,” Mr. Amodei added. “But I believe we owe it to humanity to try.”
MainStory: TopStory FederalLegislation DataSecurity AINews