ONYX has tracked the AI safety debate since September 13: Amodei’s ‘We Must Pace the Frontier’ essay; Coxon’s extinction risk warning; Altman and Musk endorsing Amodei’s three-step safety framework; Speaker Johnson urging no panic; Rep. Foster calling for regulation; Bannon at the Pro-Human Assembly; the Pentagon’s Anthropic designation reversal; and — most significantly — Senators Warner, Reed, and Coons’s September 21 letter connecting AI targeting hallucinations to the Minab school strike and a near-miss between US and Chinese forces. Today: Trump confirmed a meeting with Amodei and dismissed AI danger fears.
WHAT TRUMP SAID
President Trump confirmed a meeting with Anthropic CEO Dario Amodei. Trump repeated his view that he was not concerned about AI becoming dangerous. He warned that excessive regulation could allow China to overtake the United States in artificial intelligence.
WHAT AMODEI HAS SAID
The specific documented gap between Trump’s stated position and Amodei’s publicly stated position:
▸ Amodei: AI development is ‘one of the most transformative and potentially dangerous technologies in human history’
▸ Amodei: proposed a three-step safety framework including embedded third-party evaluators at every frontier AI lab
▸ Amodei: his CBS interview statement that he didn’t fully anticipate rapid AI’s effects was the arc’s most significant single corporate safety acknowledgment
▸ Trump (today): not concerned about AI becoming dangerous; excessive regulation could let China win
Both men met. Trump dismissed the dangers Amodei has publicly characterized as existential. ONYX covers the gap between their documented positions without asserting what was said in the private meeting.
THE AI-CHINA NEAR-MISS CONTEXT
Trump’s specific framing — excessive regulation could let China overtake the US in AI — lands in a specific documented context: seven days ago, an AI-assisted US defense intelligence system hallucinated that Chinese vessels were transporting nuclear weapons. The hallucination produced a real-world near-miss between US and Chinese forces. Senators Warner, Reed, and Coons connected AI targeting failures to the Minab strike that killed nearly 200 people.
The arc’s documented AI accountability is not hypothetical risk. It is a confirmed near-miss between two nuclear powers and a school full of children who are dead. Trump’s statement that he is not concerned about AI becoming dangerous is his documented position. ONYX covers it alongside the documented evidence that AI military systems have already produced dangerous outcomes.
THE CHINA COMPETITION FRAMING
Trump’s warning that excessive regulation could allow China to overtake the US in AI is a specific, documented policy argument. ONYX covers it at its analytical weight:
▸ The argument: if the US imposes safety regulations that slow AI development, China — which has no equivalent safety constraints — may develop more capable AI systems faster
▸ The counter-argument Amodei and others have made: unregulated AI development without safety frameworks produces exactly the kind of failure documented in the senators’ letter — a military AI hallucination that nearly started a war
▸ The specific tension: China winning an AI race is a documented concern; a US AI system nearly starting a war with China on a hallucination is a documented outcome of insufficient AI safety oversight
THE ANTHROPIC PENTAGON REVERSAL
ONYX documented on September 18 that the Pentagon rescinded its ‘supply chain risk’ designation against Anthropic — reversing a ban that had excluded Anthropic from US government contracts. Today, Trump met with Amodei. The specific sequence: Pentagon lifts Anthropic ban; Trump meets Amodei; Trump dismisses AI danger. Whether the meeting produced any specific agreements on AI governance or military AI use is from developing reporting.
Trump met with the CEO of the AI company whose safety warnings have been the most consistent in the arc. Trump said he’s not worried about AI being dangerous. Seven days ago, a US AI system hallucinated that China was smuggling nuclear weapons and nearly caused a military confrontation. Senators have formally connected AI targeting failures to a school strike that killed 200 people. Trump is not worried. Amodei has said this is one of the most potentially dangerous technologies in human history. Both men met today. One of those positions drives US AI policy.
WHAT HAPPENS NEXT
▸ AI policy — whether Trump’s stated non-concern produces any executive action limiting AI safety oversight
▸ China competition framing — whether the ‘don’t regulate or China wins’ argument shapes congressional AI legislation
▸ Military AI governance — whether the Hegseth IG investigation and senators’ letter are affected by the presidential position
▸ Anthropic government contracts — whether the Pentagon designation reversal produces specific contract awards
| CONFIDENCE: HIGH | Trump confirmed meeting Anthropic CEO Dario Amodei, who repeated not concerned about AI becoming dangerous and warned that excessive regulation allows China to overtake US artificial intelligence, per confirmed reporting. |
| ⚖️ BIAS CHECK — WHO IS SAYING WHAT | |
| Trump | Dismissing AI danger; framing as China competition issue; his position is the operative US AI policy position regardless of what safety experts say |
| Amodei | Has publicly characterized AI as potentially dangerous; proposed embedded safety evaluators; his private conversation with Trump is not available to ONYX |
| Senators Warner, Reed, Coons | Their September 21 letter documents specific AI-produced lethal and near-lethal outcomes; their evidence is directly relevant to Trump’s stated non-concern |
| China | Named as the regulatory competitor; buying 90% of Iran’s oil while simultaneously being nearly engaged in military confrontation by a US AI hallucination in the same week |
| ONYX | Covering Trump’s position and the arc’s documented AI accountability record simultaneously; not characterizing Trump’s position as correct or incorrect; naming the specific evidence that bears on it |
SOURCES
▸ Confirmed reporting — Trump Amodei AI dismissed China regulation September 28, 2026
