Wednesday, September 23: Alibaba unveils China's most powerful AI chip and plots a 10-trillion-parameter Qwen model; Xiaomi's open-weight MiMo-V2.6-Pro tops the open-source intelligence index; OpenAI publishes a framework to let outside auditors inside model training; Shanghai AI Lab drops a 744B open-weight agent; and GPT-6 Sol and Luna land on Amazon Bedrock. In this hour: Alibaba Unveils Zhenwu V900 — Triples Prior Chip, Plots 10-Trillion-Parameter Qwen · Xiaomi's MiMo-V2.6-Pro Tops Open-Weight Intelligence Index — A Phone Maker Holds the Crown · OpenAI Opens Model Training to Outside Auditors — Framework Published, Partners Not Yet Named · GPT-6 Sol and Luna Now Live on Amazon Bedrock — AWS Enterprises Get Direct Access · Shanghai AI Lab Releases 744B Atria Dawn — Biggest Open-Weight Agent Drop of the Week.

Wednesday, September 23: Trump rebrands AI as 'superintelligence' at UNGA, rejecting global oversight; Senate lawmakers revealed to be locked out of advanced AI tools they're tasked with regulating; the UNSC's first AI session closes without binding commitments as US and Chinese frontier lab heads sit together for the first time; EU AI Act enforcement formally activates full GPAI powers; and OpenAI's federal pre-emption filing draws fresh pushback. In this hour: Trump Renames AI 'Super Intelligence' at UNGA — Vows No Reining In · Senate Lawmakers Are Banned From the AI Tools They're Tasked With Regulating · First UNSC AI Session Seats US and Chinese Frontier Lab Leaders Together — No Commitments Follow · Brussels Demands Third-Party Audit of OpenAI's GPT-6 Series Under Live EU AI Act Powers.

Wednesday, September 23: Senate lawmakers exposed as blocked from advanced AI while drafting AI law; Mistral snaps up Paris adtech startup Pimento in an acqui-hire for its Vibe assistant — plus updates on the UN Security Council AI session, the EU AI Act enforcement action against OpenAI, and the Mirendil valuation surge. In this hour: Senate Lawmakers Are Banned From the AI Tools They're Trying to Regulate · Brussels Demands Third-Party Audit of OpenAI's GPT-6 Self-Assessment · First UNSC AI Session Closes With No Binding Commitments — but a New Format · Mistral Acquires Paris Adtech Startup Pimento in Acqui-Hire for Vibe Assistant.

Wednesday, September 23: The UN Security Council's first-ever joint session with frontier US and Chinese AI developers convenes on Wednesday; updates on the EU AI Act entering live enforcement, Trump's 'Super Intelligence' rebrand at UNGA, the Xi–Biden AI track at the White House, and fresh detail on OpenAI's federal standards push. In this hour: Altman, Anthropic and DeepSeek in the Room Together at the UN Security Council · Trump's 'Super Intelligence' Label Lands Awkwardly as UNSC Pushes Safety Norms · Xi Visit Yields First Formal AI Red-Lines Exchange Between Washington and Beijing · Brussels Flags OpenAI's Self-Assessment as Insufficient in First Live Enforcement Action · OpenAI's Federal Filing Adds Pre-Emption Language — Wants Washington to Override State Liability Laws.

Wednesday, September 23: Claude Opus 5.5 ships on AWS with a 40% cost cut; OpenAI opens safety evaluations to outside groups during model training; Dario Amodei publishes a 'pace the frontier' plan as Jensen Huang pushes back; the EU AI Act enters full enforcement; and OpenAI formalises a call for binding US federal AI standards. In this hour: Claude Opus 5.5 Ships — 40% Cheaper, One-Million-Token Context, Day-One on All Three Clouds · OpenAI Opens Model Training to Outside Safety Evaluators — METR and Redwood in Talks · Amodei Publishes 'Pace the Frontier' Plan — Jensen Huang Pushes Back · EU AI Act Enforcement Is Live — and OpenAI's New Evaluator Framework Maps Directly Onto It · OpenAI's Federal Standards Filing — New Detail: It Explicitly Targets State Patchwork.

Wednesday, September 23: OpenAI opens model training to outside safety auditors; Trump dismisses AI risk warnings at UNGA as the Security Council holds its AI briefing; Anthropic raises usage caps 20%; and US-China agree a formal AI incident hotline. In this hour: OpenAI Opens Model Training to Outside Safety Auditors · UNSC AI Session Convenes — Trump Calls Safety Warnings a 'Hoax' at UNGA Same Day · US and China Formally Agree an AI Incident Hotline — Shenzhen Talks in Two Months · Anthropic Raises Usage Caps 20% Across Pro, Max, and Team Plans · Altman's Eval Promise vs. Wednesday's Policy — The Gap Is the Story.

Wednesday, September 23: MIT's AI-powered microrobot hits insect-grade agility; the UN Security Council convenes its first session hosting both US and Chinese frontier AI developers; and the EU AI Act's full enforcement regime crossed its August deadline with new compliance pressure hitting companies today. In this hour: MIT's Microrobot Hits Insect-Grade Agility — 450% Speed Jump, AI Controller · UN Security Council AI Session Convenes — Altman and Chinese Labs at the Same Table · S&P: AI Adoption Will Soon Move Bank Credit Ratings · EU AI Act: Full Enforcement Is Live — No More Transition Runway for Frontier Labs.

Tuesday, September 22: OpenAI opens model training to outside safety evaluators and publishes a misalignment disclosure framework; Anthropic puts a $1B price tag on its parallel Accenture embed; the UN Security Council's Wednesday AI session seats Sam Altman and Chinese frontier-lab leaders in the same room for the first time; and AstroForge confirms Solo flies in shadow mode aboard DeepSpace-2 before going fully autonomous in 2027. In this hour: OpenAI Opens Model Training to Outside Safety Evaluators · OpenAI Publishes Misalignment Disclosure Framework — Full Lifecycle, Including Training · Anthropic Puts a $1B Price on Independent Oversight — Accenture Gets the Embed · Altman and Chinese Frontier-Lab Leaders Share a Table at the UN Security Council · AstroForge's Solo AI Flies Shadow Mode First — Full Autonomy Comes in 2027.

Tuesday, September 22: The UN Security Council convenes its first-ever session with frontier AI lab CEOs from both the US and China; Samsung takes a €3B stake in Mistral pushing its valuation to €21B; and DeepSeek joins Altman and Amodei at the Council table as the AI governance moment turns concrete. In this hour: UN Security Council AI Session Goes Live — Full Briefer List Confirmed, DeepSeek at the Table · Samsung Takes €3B Stake in Mistral — Valuation Hits €21B · Two Tracks, One Week — Bilateral US-China Pact Meets Multilateral UN Chamber.

Tuesday, September 22: Cisco Talos discloses CLOSEDQUORUM — the first documented autonomous LLM-voted malware C2 implant — while Anthropic ships Claude Opus 5.5 at 40% lower cost with new anti-distillation locks, deepening the AI price war already set off by GPT-6 Sol and Luna. In this hour: Cisco Talos Discloses First Documented Autonomous LLM-Voted Malware C2 · Anthropic Ships Opus 5.5 — 40% Cheaper, 30% Faster, With Anti-Distillation Locks.