June's AI news kept circling back to one question: who's actually in control, and what does that cost you?

Ramp's investors bet nearly $44 billion on the idea that AI spend needs its own management layer. The US government showed it can pull a frontier model out from under enterprises with only days of notice. And two very different bodies, one on governance, one on audit standards, both said board oversight of AI is overdue.

In the free section this week: why Ramp's raise matters more than another fintech funding headline, what the Fable 5 and Mythos 5 suspension actually tells us about vendor risk (including what outside critics said that Anthropic's own account left out), KPMG's new board governance principles and the ROI data behind them, and where audit tooling is racing ahead of the standards meant to govern it.

Paid subscribers get the what to do version: spend controls worth setting now, a vendor contract clause worth asking for, and this month's template, a five to ten minute standing AI agenda item for your board or audit committee, built to reuse every quarter.

This is the last week to sign up.

Four live sessions starting July 14th, on the work you actually do: variance analysis, board reporting, reconciliations, the close.

Small group, direct access to me working through your own numbers.

Outline and registration here

June's stories fall into three groups: what AI is starting to cost, who controls the models you depend on, and who is finally being asked to watch both.

Ramp raised $750 million in June at a $44 billion valuation, up from roughly $16 billion a year ago. The round was led by ICONIQ, GIC, and Ontario Teachers' Pension Plan. Ramp's pitch: AI token spend is becoming a third category of corporate spend, next to people and vendors, and it's pulling usage data from Anthropic, OpenAI, Gemini, and Cursor into a single dashboard. The company crossed $1 billion in annualized revenue with positive free cash flow.

The reason this raise landed is Uber. Its CTO told The Information in April that he was “back to the drawing board” after the company burned through its 2026 AI budget early. Bloomberg later reported Uber capped agentic coding tool spend (Claude Code, Cursor) at $1,500 per employee per month. CNBC reported a broader shift from “tokenmaxxing” to efficiency: OpenAI says average reasoning token consumption per organization grew roughly 320x over the past year, and one CEO told CNBC he moved 100% of his company's traffic to DeepSeek to cut costs.

Read more: Ramp's raise  /  the tokenmaxxing shift

So what: this is the story I'd flag first this month. Token-based billing scales with usage, not headcount, and most finance teams still don't have a line item for it, let alone a control. I wrote about why an AI bill doesn't behave like a SaaS bill a few weeks ago. That framework still holds, and it's worth a re-read now that a $44 billion valuation is riding on the problem.

Anthropic launched Claude Fable 5 and the restricted Claude Mythos 5 on June 9. Three days later, on June 12, Commerce Secretary Howard Lutnick directed Anthropic to suspend access for any foreign national anywhere, citing national security. Unable to verify nationality in real time, Anthropic disabled both models globally. The stated trigger was a reported way to bypass Fable 5's safeguards, though independent reporting describes the government's evidence as verbal and narrow: essentially, asking the model to read a codebase and fix flaws in it. Mythos 5 access returned to approved U.S. organizations on June 26, Commerce lifted the controls on June 30, and Fable 5 came back globally on July 1.

Read more: Anthropic's initial statement  /  Anthropic on restoring access  /  Forbes

So what: Anthropic's own account of this is measured, understandably. The independent reaction wasn't. AI policy analyst Dean Ball called the move “cartoonish” and said he couldn't tell if it was “lawfare against Anthropic in particular or extreme national-security hawkery.” Gary Marcus argued it worked against the government's own stated goal of staying ahead of China, warning it could push researchers back toward Chinese labs. Multiple tech executives and investors made a similar point: this handed time to Chinese open source developers.

I'd be careful about the lesson you draw here. Three days is too short for most enterprises to have built anything load-bearing on Fable 5, so this specific incident probably didn't break production workflows for most of you. The real lesson isn't “this happened to us.” It's “this can happen to any of us, on any model, with days of notice or none, over evidence a government official described as verbal.” That's the fallback-rights and vendor-concentration conversation to have with procurement now, before it's your workflow on the line.

This also connects to a bigger pattern.

A June 2 executive order created a voluntary pathway for government pre-release review of “covered frontier models,” and OpenAI's GPT-5.6 preview started with roughly 20 government-approved partners before wider release. Staged, partner-limited access to the newest frontier models may now be the norm.

KPMG, with INSEAD, released AI Governance Principles for Boards in June, pushing boards to update committee charters so audit committees own AI assurance and risk committees weigh AI's enterprise risk. The numbers behind it are worth sitting with: only 26% of boards discuss AI at every meeting. But in organizations reporting high AI ROI, 63% put AI on every board agenda, versus 13% of low-ROI organizations. Separately, ISS found only 16% of companies reviewed disclosed a director with specialized AI skills.

Read more: the governance gap

So what: the correlation between board engagement and AI returns is the most useful data point in this whole briefing. It's not proof of causation, but it's a real argument for getting AI onto the board agenda now, rather than waiting for a regulator or your auditor to ask why it isn't there.

The PCAOB opened its first-ever public comment period on its standard-setting agenda in June, including how it should consider AI in audits. The window runs through August 7. My read: this is high level and preliminary, more a signal of intent than a framework. It's good that the regulator is finally putting guidance on the table. It's also, in my opinion, a bit late given how far the tooling has already moved.

Read more: Journal of Accountancy

Meanwhile, the tooling isn't waiting for the standards. Deloitte rolled out an agentic AI network inside its Omnia audit platform for close to 85,000 audit professionals, handling risk identification, preliminary procedures, and evidence analysis ahead of human review. Workday and Microsoft both shipped governed AI agent tools for finance this month too: Workday's Agent Passport tests agents against OWASP, NIST, and MITRE standards before they touch financial data, and Dynamics 365 Copilot added AI-assisted reconciliation and natural language financial queries.

Read more: Deloitte Omnia  /  Workday  /  Microsoft Dynamics 365

So what: if agents are going to touch your books, whether it's your auditor's agents or your own ERP's, verification standards like these belong in your procurement checklist, not just your vendor's marketing deck.

We've covered this before, so just the essential: GPAI enforcement powers and Article 50 transparency duties switch on August 2, with fines up to €15 million or 3% of global turnover. High-risk uses like credit scoring were pushed to December 2027 under a pending political agreement, but the transparency duties are not deferred.

Read more: EU digital strategy

We've covered the what and the so what. In the paid section, we get to the what to do: spend controls, vendor language, and audit prep you can act on this week. The deliverable of the month is the AI Standing Agenda for the Board, a five- to ten-minute board brief you can drop straight into your next meeting and reuse every quarter after.

Thanks for reading this one all the way through. It was a dense month.

Here's what I'd want you to sit with: June's headlines were about spend discipline and board oversight catching up to what's already been deployed, not about a model getting any smarter. That's the actual work in front of most finance teams right now.

See you next Tuesday.

Anna

This newsletter is for you, and we want to make it as valuable as possible. Please reply to this email with your questions, comments, or topics you'd like to see covered in future issues. Your input shapes our content!

Want to dive deeper into balanced AI adoption for your finance team? Or do you want to hire an AI-powered CFO? Book a consultation!

Did you find this newsletter helpful? Forward it to a colleague who might benefit!

Until next Tuesday, keep balancing!

Anna Tiomina
AI-Powered CFO