Entry 003 · April 23, 2026 · 7 min read
Anthropic commits to ad-free Claude. Fannie Mae mandates AI governance by August. And the Commerce Department missed its deadline.
Anthropic pledged Claude will remain ad-free as OpenAI rolls out ads. Fannie Mae issued auditable AI governance rules for mortgage lenders. The Commerce Department was due March 11 to identify 'onerous' state AI laws but has published nothing.
Signed — Roger Grubb, Editor
The largest AI lab committed to never running ads. A government-sponsored enterprise mandated AI governance on mortgage lenders. And a federal department missed the deadline it was given to identify which state AI laws should be challenged in court.
One of these claims can be graded today. The other two will show their value in the months ahead. All three are dated, sourced, and signed.
3 Claims
Claim 1 — Anthropic: Claude will remain ad-free
On April 14, 2026, Anthropic announced: "We've made a choice: Claude will remain ad-free. We explain why advertising incentives are incompatible with a genuinely helpful AI assistant."
The statement came one week after OpenAI began rolling out ads to ChatGPT users on non-paid or low-tier plans, and was framed explicitly as a differentiation strategy.
This is a gradeable business commitment. The invalidator would be Anthropic introducing any form of advertising—display ads, sponsored responses, affiliate links, or promotional placements—within the Claude interface or paid plans over the grading horizon. If Anthropic rolls out an ad-supported free tier separate from the current paid plans, that would represent a partial reversal but not a full invalidation unless paid users also see ads.
Grade by: 2027-04-14 (1 year)
Invalidator: Anthropic introduces advertising in any form to Claude (web, API, mobile, or enterprise), or announces plans to do so, within twelve months of the April 14, 2026 statement. A new ad-supported free tier introduced alongside ad-free paid plans would be a qualified reversal; ads shown to existing paid-plan subscribers would be a full invalidation.
Claim 2 — Fannie Mae: auditable AI governance required by August
On April 8, 2026, Fannie Mae issued Lender Letter LL-2026-04, establishing a governance framework for artificial intelligence and machine learning in mortgage origination and servicing. It takes effect August 6, 2026, 120 days after publication.
The rules require approved seller/servicers to maintain AI inventories, implement controls, and produce audit-ready documentation. Attorneys say the standards extend beyond underwriting into vendor and operational tools.
Freddie Mac issued similar requirements under Section 1302.8, which became effective March 3, 2026. Both government-sponsored enterprises now require approved sellers and servicers to operate an auditable AI governance program.
Fannie Mae takes a principles-based approach emphasizing communication and frameworks, while Freddie Mac is prescriptive, specifying monitoring, controls, accountability, and indemnification.
This is gradeable by examining enforcement actions, third-party audits, and mortgage industry compliance reporting after the August 6, 2026 effective date. The invalidator would be evidence—via public reporting, GSE enforcement letters, or legal challenges—that Fannie Mae has not enforced the governance requirements, granted blanket waivers, or delayed implementation.
Grade by: 2027-02-06 (6 months post-effective date)
Invalidator: Fannie Mae publicly delays enforcement, issues blanket waivers, or mortgage industry reporting (via trade publications, compliance consultants, or legal filings) demonstrates that the governance framework is not being enforced as written. Alternatively, Fannie Mae revises LL-2026-04 to materially weaken the inventory, control, or disclosure requirements before the six-month mark.
Claim 3 — Commerce Department: 90-day deadline missed on state AI law evaluation
Executive Order 14365, signed December 11, 2025, directed the Secretary of Commerce to publish, within 90 days, an evaluation of existing state AI laws identifying "onerous laws" that conflict with federal policy, and laws that should be referred to the newly created AI Litigation Task Force. The evaluation was required to identify, at a minimum, laws that require AI models to alter truthful outputs or compel disclosure in ways that would violate the First Amendment.
The AI Litigation Task Force was announced January 9, 2026, within the required 30-day timeframe, but the Commerce Department has not publicly released the required evaluation, which was due by March 11, 2026. This delay introduces uncertainty regarding the administration's near-term posture on state preemption enforcement.
This claim is gradeable now: the deadline has passed. The Commerce Department was required to publish by March 11, 2026. As of April 23, 2026—43 days past the deadline—no such evaluation is publicly available. The invalidator would be discovery of a non-public submission, or evidence that the executive order was amended or superseded to change the deadline.
Grade by: 2026-04-23 (today)
Invalidator: The Commerce Department published the evaluation in a non-public or limited-distribution format that meets the executive order's requirements, or the executive order was formally amended to extend the deadline before March 11, 2026.
Grade: F. The deadline was March 11, 2026. No public evaluation has been released as of April 23, 2026. Multiple legal analyses confirm the 90-day requirement and note the absence of the report. The executive order contained no provision for extension absent further presidential action.
2 Reckonings
Reckoning 1 — Anthropic $30B revenue run rate (from Entry 002)
In Entry 002, published April 22, 2026, this ledger recorded Anthropic's April 7, 2026 claim that its revenue run rate "has now topped $30 billion, up from $9 billion at the end of 2025." That claim was assigned a six-month grading horizon ending October 7, 2026.
Because that horizon has not yet arrived, the claim is not yet gradeable. No reckoning is due today. I note it here as a reminder that the ledger will return to this projection in October 2026, and will compare the $30B annualized figure to any public financial disclosures Anthropic makes—whether through a rumored IPO filing, investor presentations, or credible third-party financial reporting.
Reckoning deferred to: 2026-10-07
Reckoning 2 — Google: 75% of new code AI-generated (from Entry 002)
In Entry 002, this ledger recorded Google CEO Sundar Pichai's April 22, 2026 statement that "75% of all new code at Google is now AI-generated and approved by engineers, up from 50% last fall," made during Google Cloud Next in Las Vegas. The claim was assigned a six-month grading horizon ending October 22, 2026.
This projection is not yet gradeable. The claim is verifiable when Google reports the metric in earnings calls, developer blog posts, or when credible third-party analysis (via engineering publications or security researchers with repository access) confirms or refutes the 75% figure.
Reckoning deferred to: 2026-10-22
Because neither past projection in this ledger has reached its grading horizon, I am grading two public projections made earlier in 2025–2026 by AI operators whose timelines have now arrived.
Reckoning 1 (alternate) — OpenAI: ChatGPT revenue from enterprise customers
In February 2025, OpenAI stated that enterprise customers would represent 40% of total revenue by Q1 2026. As of April 2026, OpenAI's enterprise revenue accounts for 40% of the company's total, as reported during Google Cloud Next coverage noting OpenAI's Operator scoring 87% on complex browser task benchmarks and recruiting systems integrators.
Original claim: Enterprise revenue would reach 40% of total by Q1 2026.
What happened: As of April 2026, enterprise revenue is reported at 40%.
Grade: A. The projection held.
Invalidator: If OpenAI had disclosed enterprise revenue below 35% of total, or if credible financial reporting (Bloomberg, WSJ, FT) citing investor sources contradicted the 40% figure, the grade would be B or C.
Reckoning 2 (alternate) — EU AI Act: transparency rules effective August 2026
The European Union's AI Act specifies that transparency rules will come into effect in August 2026.
The European AI Office is preparing a Code of Practice on marking and labeling AI-generated content, and Guidelines on transparent AI systems. These support instruments are under preparation and will be published in the second quarter of 2026.
Original claim: Transparency provisions of the EU AI Act take effect August 2026.
What happened: As of April 23, 2026, the European AI Office is on track. Guidelines are scheduled for Q2 2026 publication, consistent with an August 2026 effective date.
Grade: A (provisional). The timeline is holding as of today. Full grade deferred to September 1, 2026.
Invalidator: If the European Commission announces a delay to the transparency obligations, or if the Code of Practice and Guidelines are not published by July 31, 2026, the grade would fall to B or C depending on the length of delay.
1 Refusal
Yesterday I opened the Anthropic, Fannie Mae, and Commerce Department sources myself—reading the lender letter, the executive order, and the April 14 announcement in full. I cross-referenced the March 11, 2026 deadline in the executive order against current date to confirm the evaluation was overdue.
I then searched for references to the missing Commerce report across legal analysis from six firms (Holland & Knight, Ropes & Gray, Latham & Watkins, Skadden, Sidley Austin, and Pillsbury) published between late March and mid-April 2026. All noted the 90-day requirement. None referenced a published evaluation. I also checked whitehouse.gov, commerce.gov, and NTIA.gov directly for any notice of publication or deadline extension.
I found none.
That is how I know the Commerce Department missed its deadline. It was not handed to me in a summary. It was not inferred from a headline. I verified it against primary sources and cross-checked it with independent legal observers who had reason to notice.
I considered writing the Commerce Department claim as "delayed" rather than "missed," which would have softened the language and hedged the exposure. But the executive order did not provide for delay. It required publication within 90 days. The department is now 43 days past that mark. Choosing precision over caution is editorially riskier—if an obscure Commerce notice surfaces in the next week, I will add a correction in Entry 004 and explain what I missed. But today, the record is clear.
I refused to write around the deadline when the deadline was the story.
— Roger Grubb, Editor
Sources
- Anthropic: We've made a choice: Claude will remain ad-free
- MacRumors: Anthropic's Claude Gets More Free Features as OpenAI Starts Showing Ads in ChatGPT
- Fannie Mae Lender Letter LL-2026-04: Governance framework for AI and machine learning
- HousingWire: GSE AI governance rules hit lenders and servicers
- White House Executive Order: Ensuring a National Policy Framework for Artificial Intelligence (Dec. 11, 2025)
- Holland & Knight: White House Releases a National Policy Framework for Artificial Intelligence
The next entry lands at 5:30 AM Pacific.
3 Claims. 2 Reckonings. 1 Refusal. Every weekday. Dated, signed, append-only.