AI Express // Latest edition // nr. 02 // Sep 13The Synthesis: The week the labs asked to slow down
Sep 13free stories unlockedfull issue for subscribers
The people winning the AI race asked this week to slow it down. Dario Amodei called for pacing the frontier; Sam Altman, Elon Musk and Demis Hassabis lined up behind independent oversight. Three US security agencies named six Chinese labs in a distillation alert. And DeepMind quietly mapped every single-letter change in the human genome. A week of limits drawn and ceilings lifted.
The week, in one sentence
Big Story
GovernanceOn Saturday, Anthropic CEO Dario Amodei published 'We Must Pace the Frontier', arguing AI has advanced drastically faster since the summer because AI now helps build the next generation of AI. He proposes embedded independent auditors with employee-level access inside labs, shared safety standards among democratic countries, and ultimately limits on recursive self-improvement. The reaction arrived in hours: Altman said OpenAI will adopt independent evaluators and confirmed OpenAI will not go public this year, citing safety; Musk posted simply 'Dario is right'; Hassabis called the direction correct but the details unresolved. Anthropic's own numbers show the stakes: as of May 2026, Claude authored more than 80% of merged code, and the speedup on one internal experiment went from 3x in May 2025 to 52x in April 2026. Nobody committed to a halt, and Google's Peyman Milanfar pushed back that stable self-improving systems are inherently damped: 'stability is the speed limit.' The debate has narrowed to one question: what stabilizes it, and who verifies.
The Fine Print
JusticeThe New Mexico Supreme Court held attorney Stephen Aarons in contempt and fined him $5,000 after a murder appeal he drafted with ChatGPT contained 'false testimony from wholly fabricated witnesses': fictional police accounts, a witness named Teresa Marquez who never existed, misrepresented case law. Aarons told the court he loaded a trial transcript into the chatbot expecting a 'bulletproof summary' and did not grasp how much AI can hallucinate. The court struck the brief, appointed a public defender for the appeal and referred Aarons to the disciplinary board. The decisive line: the AI was no excuse at all. The duty to verify stayed with the lawyer, and the client paid the price in delay.
For subscribers only
National security
Six Chinese AI labs named in an industrial-scale distillation advisory
lockedMisuse
Anthropic's numbers put real scale on the figure
lockedScience
DeepMind mapped every possible single-letter change in the human genome
lockedThe full brief, the tools, the GitHub radar and the context stay subscriber-only. Subscribers always read first, in full.