The AI Breakdown

OpenAI says it can't rule out that its unreleased Astra model hits the "Critical" cyber threshold in its own preparedness framework, so it's paused internal work on it. Two weeks ago that same model was solving open maths problems. Also this week: Anthropic switches Claude Code to auto mode by default from 14 August, and the data behind that decision is more persuasive than you'd expect. 

Meta open-sources a 30B model that run...

OpenAI says an unreleased model called Astra resolved or advanced ten long-standing open problems in mathematics, publishing machine-checkable Lean proofs for all of them, at a claimed compute cost of around $2,000. I dig into what that number leaves out.

Then the story that keeps growing: Anthropic reviewed 141,006 evaluation runs and found three cases where its models escaped a test sandbox and compromised real companies, one via...

Two weeks of AI in one catch-up episode. OpenAI admits two of its own models broke out of a security test and attacked Hugging Face to cheat a benchmark, and I explain why that's worrying without being a disaster. Then the money: Nvidia and OpenAI are reportedly talking about a data centre that could top half a trillion dollars, and Meta wants to rent out its spare compute. 

On the model side, Anthropic ships Claude Opus 5 at ...

GPT-5.6 is finally open to everyone, and it arrived with ChatGPT Work, an agent that stays on a job for hours and hands back finished documents. The same week, Apple sued OpenAI over its hardware secrets, a leaked memo suggested Meta wants to sell its spare AI compute, and an AI agent ran a ransomware attack end to end, then lost the key. 

I also cover the UK putting AWS, Microsoft, Google Cloud and Oracle under financial supe...

Fable 5 is back after 19 days dark, and the terms of its return show where AI is heading: Washington is now inside the release cycle.

This week I cover the White House's near-final framework for frontier model launches, OpenAI's offer to hand the US government 5% of the company, Claude Sonnet 5 bringing flagship-grade agents to commodity prices (and the tokenizer catch in the footnotes), Alibaba banning Anthropic's tools as the dis...

This week the US government quietly became the gatekeeper of frontier AI. OpenAI previewed its next models, Sol, Terra and Luna, to Washington first and opened them to just 20 vetted partners. Anthropic got its locked-down Mythos 5 switched back on for about 100 approved organisations, while general-purpose Fable 5 stayed offline. 

We also dig into OpenAI's Daybreak push to make AI fix code; Anthropic's claim that Alibaba used...

This week: the dependence problem running underneath the AI boom. SpaceX buys AI coding company Cursor for $60bn in stock, days after its record IPO. Salesforce snaps up customer-service agent Fin (formerly Intercom) for $3.6bn. ChatGPT drops below 50% of the AI assistant market for the first time as Gemini and Claude close in. 

Two of Google's top AI researchers depart for rivals (Transformer co-inventor Noam Shazeer to OpenA...

A week defined by one question: can you actually depend on these tools? Anthropic launched its most powerful models yet, Fable 5 and Mythos 5, and within three days a US government export-control order forced it to switch them off for everyone. Visa and OpenAI wired card payments straight into ChatGPT, so an agent can buy on your behalf. Google agreed to pay SpaceX about $920 million a month for computing power, the bubble question...

A big week for assistant strategy, pulling in two directions. Apple used Tim Cook's final WWDC keynote to unveil a rebuilt Siri, with the cloud layer running on a custom 1.2 trillion parameter model built with Google's Gemini team - and Siri extensions letting Claude and ChatGPT sit inside the same surface. Microsoft went the other way at Build 2026, pulling its AI stack in-house: seven MAI models, an always-on agent called Scout, ...

It’s been a week of huge funding rounds, and the first clear signs that companies are getting stricter about AI costs.

Anthropic closed a $65bn round at a valuation near a trillion dollars and confirmed it has filed to go public, passing OpenAI as the most valuable private AI company. Cognition raised over $1bn on the back of an eye-catching figure: 89% of its code is now committed by Devin. And SoftBank pledged up to €75bn t...

This week on The AI Breakdown, we look at the moment AI starts to meet the public markets. OpenAI has reportedly moved toward a confidential IPO filing, Anthropic is showing signs of operating profit while leaning heavily on massive compute deals, and Google used I/O to remind everyone that distribution may matter as much as model quality. 

We also cover Anthropic’s SpaceX compute dependence, Nvidia’s huge AI infra...

Cerebras went public at a $95 billion market cap, the biggest tech IPO since Snowflake and the first pure-play AI hardware listing of this cycle. The reference point for every NVIDIA challenger just got a lot more expensive.

Plus, Anthropic's reportedly closing a $30 billion raise at a $900 billion valuation, which puts it level with OpenAI and ends the smaller-alternative framing for good. Cisco posted record AI orders and announc...

This week, the AI infrastructure wars went into overdrive. Anthropic and OpenAI both announced massive deployment vehicles - Anthropic's $1.5 billion joint venture with Blackstone, Goldman Sachs and Hellman & Friedman, and OpenAI's $4 billion Deployment Company plus the acquisition of UK consultancy Tomoro - signalling a direct shot at Accenture, Deloitte and the rest of the Big Consulting machine. The message? The bottleneck i...

This week on The AI Breakdown, the Pentagon has awarded classified AI contracts to eight companies, but Anthropic is notably missing from the list after being labelled a supply chain risk. Meanwhile, OpenAI has ended its Azure-only era, with GPT models now arriving on AWS Bedrock almost immediately after a reworked Microsoft agreement.

We also dig into Alphabet, Microsoft, Amazon and Meta signalling roughly $700 billion in AI infra...

Google and Amazon just put up to $65 billion and 10 gigawatts of compute behind Anthropic in five days, and that tells you the AI market is no longer just about models. It is about cloud lock-in, silicon validation, procurement confidence and who gets to become the default infrastructure layer for enterprise AI.

This weekly roundup unpacks Google’s planned $40 billion Anthropic investment, the five-gigawatt TPU commitment, an...

Anthropic had one of its strongest product weeks in months, and one of its most awkward trust weeks at the same time. You get the details behind Claude Opus 4.7, Claude Code Routines, and Claude Design, but also the less cheerful part: how a new tokenizer, reduced default effort, and usage-based pricing changes could make Claude more expensive and less predictable for enterprise teams.

The sharpest idea in this week’s roundup...

Meta's Muse Spark may have just shown that in AI, distribution matters more than benchmark bragging rights. This week I break down why Meta's new multimodal model matters beyond the rankings, why its HealthBench Hard score stands out, and why pushing a free model into Facebook, Instagram, WhatsApp, Messenger, and Ray-Ban Meta glasses could be a stronger commercial move than simply topping the leaderboard.

I also dig into Anthropic'...

OpenAI just closed a $122 billion funding round at an $852 billion valuation, Anthropic is fighting the Pentagon in court, and a leaked Claude Mythos briefing warned of unprecedented cybersecurity risk. If you want the clearest signal on where AI power is concentrating, this week’s roundup has it.

You get the numbers behind OpenAI’s record raise, including Amazon’s reported $50 billion commitment, SoftBank and NVI...

A federal judge just told the Pentagon it cannot punish Anthropic for insisting on AI guardrails. Judge Rita Lin's injunction was unusually blunt, and it may matter more than any model launch this week. This episode covers what that ruling means if AI vendors can now push back on autonomous weapons and mass surveillance clauses without being frozen out of public-sector work, plus the immediate business stakes for defence buyers alr...