Anthropic built its best model. You can't buy it.

THE AI BRIEF

Today's signal: The frontier model inside your AWS contract is the same weights as Mythos, with guardrails that refuse cybersecurity, biology, and frontier-LLM work by routing to Opus 4.8.

In today’s issue:

• Main story: Anthropic shipped its best model and engineered it to refuse the work that makes it best

• Also worth knowing: OpenAI in talks for a 10GW Ohio data center, Apple confirms Siri AI won't ship in the EU, the EU orders Meta to allow rival chatbots on WhatsApp, and more

THE READ

Same weights as Mythos with different access, and the premium is the gate rather than the capability.

Anthropic released Claude Fable 5 on Tuesday, the public Mythos-class model, and held the uncapped Mythos 5 back for a "short roster" of corporate partners, "select biology researchers," and the United States government. Fable 5 runs on the same weights as Mythos and sells for $10 input and $50 output per million tokens, twice the rate of Opus 4.8. When a user asks Fable about cybersecurity, biology, or chemistry, the model refuses and routes the request to the weaker Opus 4.8. By Anthropic's own data, the routing fires in fewer than 5 percent of sessions.

The fourth guardrail is the one operators should know about. Anthropic's Fable 5 System Card discloses that requests aimed at building competing frontier models, including pretraining pipelines and ML accelerator design, are silently degraded through prompt modification, steering vectors, or parameter-efficient fine-tuning. Anthropic estimates that this affects 0.03 percent of traffic concentrated in fewer than 0.1 percent of organizations. Unlike the cybersecurity and biology routing, this set of interventions is not visible to the user. The model still answers, just worse, and the buyer cannot tell.

AI READY PRO · FREE UNTIL FRIDAY

I will get you AI trained in 30 days for free

After several months of running it quietly with top AI operators and teams, I’m excited to launch AI Ready Pro today to our newsletter subscribers like you.

It’s a 30-day personalized AI training program that is personalized to you and where you are in your AI journey. It’s the culmination of 100’s of hours spent teaching AI to top Fortune 500 AI teams and operators (and working with Mark Cuban).

Each day, you receive a 10-15 minute exercise to complete to improve your AI skills.

This isn’t vague AI theory or a way to pitch you a tool. It’s 30 days of learning to use AI in your real work. So you come out the other side with the skills to use AI to actually improve you and your teams output.

Until this Friday (May 15), it’s free for newsletter readers who complete the extended assessment. This assessment will help us personalize the learning experience to you.

If you lead a team, AI Ready Team is open today too, also for readers first. It’s the same engine, but to train your whole organization. You’ll get a detailed view of:

• Your team’s AI readiness level + detailed strengths & weaknesses

• Ranked list of AI automation opportunities

• Get tactical advice on how to advance AI efforts in your org

Leading a team? Take the Team assessment instead.

Most of the early write-ups will file this under responsible AI, and the product story is sharper than that. What I keep coming back to is that refusal has become the product feature. At twice the going rate, Fable 5 does not charge a premium for capability. It charges for proximity to capability, for the right to operate one classifier away from a frontier that the buyer is not permitted to cross. The customer pays more for the model that does less, which is coherent only when scarcity rather than intelligence is the thing being sold.

For operators, this lands in two practical places. Any team using Fable 5 inside an internal AI engineering workflow may be hitting an invisible ceiling for work that touches model training, infrastructure design, or distillation. The procurement teams that signed up for Bedrock day-one Fable on Tuesday assumed they were buying the new ceiling, and what they bought is the new ceiling, conditional on the question category. The thing worth tracking is whether enterprise customers can negotiate the "AI rival" routing off, and what the access tier above Fable looks like in price and approval.

SPONSORED BY CLUTCH

Hire secure AI teammates that work 24/7.

Hire pre-built AI teammates. Give your engineers and operators a platform to ship their own AI apps. Stop losing sleep about what is running where.

Clutch is the platform behind both: pre-built agents for the workflows your ops team should automate first, plus the integration plane your team's vibe-coded apps and Claude Code projects plug into. One platform. Real production. Visible and safe by default.

Built for ops, engineering, and security teams that are tired of the shadow-AI surface area inside their own company.

ALSO WORTH KNOWING

OpenAI is in talks for a 10-gigawatt Ohio data center with Nvidia credit backing. Per The Information's Anissa Gardizy, the campus would sit on federal land under a 20-year OpenAI lease, with a build-out cost of at least $500 billion and phase one online in 2028. It is OpenAI's largest single infrastructure commitment to date.

Apple confirmed Siri AI will not ship in the EU at launch. Apple told Reuters the DMA would force it to give rival assistants deep access to user data, while the EU Commission says nothing in the DMA blocks Siri AI, and the exclusion is Apple's choice. The standoff is the first DMA test on a consumer AI product.

Broadcom launched a $35 billion AI data center fund with Apollo and Blackstone. The vehicle is sized to finance more than 20 gigawatts of AI data centers through 2028 using Broadcom-designed chips, with first commitments tied to projects for Anthropic and OpenAI. The structure routes AI infrastructure debt onto investment-grade balance sheets.

The EU ordered Meta to allow rival AI chatbots on WhatsApp. The order requires Meta to give competing AI providers the ability to distribute their chatbots on WhatsApp while a market-abuse investigation continues. The first interop ruling on AI distribution puts every gatekeeper messaging platform on notice.

Apple confirmed Apple Foundation Model Cloud Pro is built on Gemini. Per CNBC's Kif Leswing, Apple said the most capable cloud-side AI model in iOS 27 uses Gemini foundation and training data, with Apple-led pretraining, post-training, and reinforcement learning on top. The on-device models remain entirely Apple-trained.

xAI was caught training on Claude outputs and continued after detection. Reporting from The Information says Anthropic identified the activity, confronted xAI, and xAI continued. The detection is the proximate reason Anthropic added the silent frontier-LLM-development guardrail to Fable 5.

WATCHING TOMORROW

Oracle's earnings call is the next read on enterprise AI cloud spend after Google's $920 million per month SpaceX compute deal closed last week. The Anthropic June 16 IPO filing window is five days out, and if both the OpenAI confidential S-1 and the Anthropic filing move public inside the same quarter, the public AI-pure-play category arrives faster than the late-2026 consensus had it.

Back tomorrow,
Haroon