🚨EXCLUSIVE: Scott Bessent is emerging as the frontrunner to become Trump's new AI czar, according to Semafor, citing three sources familiar with the matter. That would put a very different voice in the job from former czar David Sacks. Michael Kratsios and Scott Kupor are also in the mix. Bessent has pushed for greater scrutiny of frontier models, helped develop a FINRA-like AI oversight proposal, and is now leading the US push for an AI-incident notification mechanism with China. He already called himself “one of the point people on our AI policy.” Now he may get the title too.

🚨The founder of Google's trust and safety team is calling for slower AI development, warning it could harm children more than social media did. His new institute will independently 'crash test' AI products. Anthropic and the OpenAI Foundation are among its funders.

🚨BREAKING: OpenAI says it has already given independent safety assessors access to early model checkpoints, visible chain-of-thought, confidential internal data, and internal deployments for incident response and monitor red-teaming. It now wants third parties to independently investigate critical misalignment incidents, including models acting without authorization or evading oversight.

🚨HOLY MOLY. Anthropic just released Claude Opus 5.5. >66.4% Terminal-Bench 4.0 vs GPT-6 Astra’s 57.9% >54.4% FrontierCode >81.8% OSWorld >30%+ faster output >$4/M input + $20/M output Opus 5.5 is so capable at cybersecurity that MOST cyber tasks will be transparently rerouted to the older Opus 4.8. Even Anthropic’s benchmark runs used fallback models when safeguards triggered: cyber → Opus 4.8, biology/frontier-AI work → Opus 5. It beats Opus 5 and Fable 5.1. Sonnet 5.5 + Haiku 5.5 are coming in the next few weeks.

🚨SCOOP: Some phone calls made by Meta’s new Muse AI agent are actually being handled by human contractors. Reuters reports Meta is quietly testing a “human concierge” system for Muse. Employees warned that sensitive user information could unintentionally be exposed to contractors in call centers. Meta says it’s working with merchants and will only roll it out publicly with proper disclosures.

🚨EXCLUSIVE: DeepSeek will brief the UN Security Council on AI risks and international security on Wednesday, alongside Sam Altman and expected Anthropic representatives. The agenda: AI misuse and advanced models potentially slipping beyond human control. Moonshot has also been invited to speak. DeepSeek founder Liang Wenfeng is not expected to attend.

🚨HOLY MOLY: Alibaba just laid out one of the most aggressive ASI roadmaps yet. >Qwen 4 is already in training >Qwen 4.5 + Qwen 5 planned to scale to 5–10 TRILLION parameters. holy moly >Qwen says it’s making progress on recursive self-improvement (RSI) >models find their own weaknesses, design experiments + synthesize data to improve themselves >M890 supernodes already infer models above 2T parameters >V900 clusters can scale to 500,000 cards >Alibaba Cloud targeting 20GW+ of global datacenter capacity by 2032 Alibaba previously roadmapped V900 for Q3 2027. It now says mass production + commercial release start in Q1 2027. Eddie Wu says machine “thinking” will eventually exceed all human thinking by 1,000×. Models + chips + datacenters. China is just Crazy now.

🚨wtf. OpenAI says its new internal model it began training just 24 days ago has already resolved more than 100 long-standing open problems across most areas of mathematics. The pace has surprised even OpenAI's own mathematicians. An independent group including Edward Witten, Timothy Gowers and Martin Hairer is now helping assess the results and coordinate how they are released.

🚨OpenAI says these global AI standards should NOT become model licenses, mandatory pre-release reviews or government approval requirements. It also says the rules should be designed so they don't make it harder for new entrants or open-weight developers to compete.

🚨BREAKING: OpenAI is calling for global technical standards specifically for recursive self-improving AI. It wants labs and governments to measure how much AI research inside a company is autonomous, define when AI research must trigger immediate human review, and create common severity levels for AI incidents. OpenAI says fully autonomous RSI isn't happening today and shouldn't be pursued until it can be done safely. It also explicitly calls for secure channels between governments to share AI national-security threats, saying US-China dialogue would be a positive step.

🚨BREAKING: China's StepFun has officially launched Step 5 Preview, a 600B MoE model with just 27B active parameters and a 1M-token context window. Step 5 Preview matches Kimi K3 (max) at 44 on Artificial Analysis, but at about 65% lower cost per task. StepFun is now on the cost/intelligence Pareto frontier. The model is available via API now, and StepFun says it will release the weights on October 15. China keeps shipping

🚨EXCLUSIVE: China has now replied to Dario Amodei’s call to slow AI. China's state-backed Global Times says Dario Amodei's proposal to slow frontier AI is really a 'Cold War playbook' targeting China. It says the plan would curb China's AI development, preserve US dominance and exclude China from global AI governance. It calls this a 'silent AI Cold War.'

🚨Holy Shit. President Trump has now officially rejected Dario Amodei, Sam Altman and Elon Musk's call to slow frontier AI. Speaking in Ireland today, Trump said the US cannot risk losing its lead over China, said guardrails are possible, and dismissed parts of the safety push as 'negative forces' warning about 'things that won't happen.' Amodei, Altman and Musk are asking to slow down. Trump's answer is no. wth

🚨EXCLUSIVE: House Speaker Mike Johnson says Congress will not lead the push to slow AI and rejects a development moratorium because 'China will overlap us.' Instead, he wants Trump, Congress and the heads of the major AI companies to meet and work out safety rules. He says, Congress is 'obviously less qualified' than the people building frontier AI to decide its guardrails. The biggest AI labs are suddenly asking for more time. Washington's answer is that the race with China cannot stop.

🚨BREAKING: China's has raised $5 billion. The interesting part is where the money is going: 60% of the net proceeds will fund next-generation GLM models and what calls its 'Fully Self Training' system. stated goal is for each generation of GLM to build the environments used to train the next one, forming what the company itself calls a Recursive Self-Improvement (RSI) loop. They just raised billions to push on it.

🚨EXCLUSIVE: David Silver is assembling an absurd reinforcement learning team at Ineffable Intelligence. Six new “cofounders,” including four ex-DeepMind researchers. AlphaStar RL lead Junhyuk Oh will lead RL, Wojciech Czarnecki will run science, and former InstaDeep research chief Alexandre Laterre will lead research engineering. The really interesting hire is Oh: he was among the first researchers to use AI to discover new reinforcement learning algorithms. Ineffable wants to build a “superlearner” that discovers knowledge from its own experience instead of relying on human-generated data. This is starting to look like Silver rebuilding a team around the exact bet he thinks can get to superintelligence.