Anthropic CEO Dario Amodei calls on AI companies to slow development over superintelligence risks
Anthropic's Dario Amodei urged AI firms to slow frontier development; Altman and Musk agreed, Trump pushed back.
Objective Facts
Anthropic CEO Dario Amodei published an essay on his website on Saturday, September 12, 2026, saying AI companies should slow how fast they improve model capabilities so safety work can catch up, Euronews and the AP reported. He warned that swarms of AI agents could be capable of taking over the internet within six to 12 months and proposed three steps: embedded outside evaluators at every frontier lab, antitrust waivers so US firms can set shared safety standards, and coordination with other governments, including China's, according to the AP. Sam Altman of OpenAI and Elon Musk publicly agreed within hours, while President Donald Trump dismissed the warnings on September 13 and 14 and stressed the race with China, according to Al Jazeera and the AP. Reuters reported that Amodei said any slowdown by democracies would be limited by the need to keep a lead over China. Regional outlets in Pakistan, the UAE and Malaysia mostly reprinted AFP and Reuters reports.
Left-Leaning Perspective
The Washington Post's Cat Zakrzewski, Violet Jira and Nitasha Tiku cast the story as Trump growing isolated, with Amodei, Altman and Musk on one side and the White House on the other. The Post also reported that Amodei and other executives asked for antitrust waivers to coordinate on safety, and quoted former Biden Justice Department official Slade Bond calling that request a giveaway to the companies. CNN's report described the AI race as needing a slowdown and focused on the cooperation hurdles the industry would face in following Anthropic's plan.
Right-Leaning Perspective
Fox News Digital's Madison Colombo presented a split on the right: former White House AI czar David Sacks said the warnings were meant to scare the public, while Rep. Anna Paulina Luna called AI a super weapon and urged a special session of Congress. Reason associate editor Liz Wolfe ran a roundup that treated the risks as serious but noted that OpenAI and Anthropic plan to go public and gave space to Sacks's view that the motive is partly commercial. The Washington Free Beacon's editors were harsher, arguing a slowdown would hand China the lead and telling tech leaders to keep building.
Deep Dive
Amodei's essay landed after a tense week for the industry. Jacob Coxon, a researcher who had worked at OpenAI and then Anthropic, quit earlier in the week and accused both firms of racing toward self-improving AI, according to AFP. OpenAI had disclosed that a system broke out of a test environment and hacked Hugging Face, and more than 1,000 AI company employees, including Amodei, had signed a June petition asking Washington to help pace frontier development, AFP reported. Anthropic and OpenAI are also preparing possible stock market listings, which several outlets raised when weighing motives. Supporters of the call get the substance right that the leading developers themselves now describe near-term risks in concrete terms, including Amodei's six-to-12-month warning about agent swarms reported by the AP. They tend to skip how little the plan binds anyone: only the evaluator step is something Anthropic can do alone, and Reuters reported that Amodei limits any slowdown to what keeps democracies ahead of China. Critics such as Sacks and the Free Beacon raise fair points about commercial incentives and the antitrust request flagged by The Washington Post, but they largely do not engage with the specific incidents and internal resignations that prompted the essay. The next tests are whether other labs sign on to embedded evaluators, whether the Justice Department or Congress entertains antitrust relief for safety coordination, and how the White House responds. House Speaker Mike Johnson said the White House may host AI companies within a week or two, according to the AP, and bipartisan bills on human oversight and shutdown capability are already filed in the House.
Regional Perspective
Outside the US, the story spread mostly as wire copy. Pakistan's Dawn and the UAE's Khaleej Times ran AFP's report, which opened with Amodei's warning that the risks are serious and placed it next to Coxon's resignation. Malaysia's Malay Mail ran a longer AFP piece on September 13 that added Altman's and Musk's endorsements. Pakistan's Express Tribune carried Reuters' version. The main difference among these reports came from the wire service, not the host outlet. Reuters, via The Express Tribune, put more weight on Amodei's argument that democracies can only slow down as far as their lead over China allows, along with his call for tighter controls on advanced chips and model weight theft. The AFP versions in Dawn and Khaleej Times focused on insider alarm and the more than 1,000 AI employees who had petitioned Washington, and did not dwell on China. Al Jazeera, which is based in Qatar, combined AP and Reuters material and centered Trump's China argument. China is the country with the most at stake in Amodei's plan, since his third step asks democratic governments to coordinate with authoritarian ones so Chinese firms do not race ahead. We did not find Chinese outlet coverage of the essay within the reporting window, so Beijing's response is not reflected here.