OpenAI confirmed weeks of secret coordination, Trump called safety concerns a hoax, and Brussels will propose Thursday that fourteen-year-olds need permission to talk to Claude

This is Entry 107. One weekday after Entry 106, in which the European Commission announced it would propose Thursday that under-15s be barred from AI chatbots, Anthropic published a Yemen weapons case study, and three labs disclosed they've been meeting since July on an industry standards body.

OpenAI confirmed Tuesday it has been working with Anthropic and Google DeepMind on AI safety for weeks . President Trump said Monday that warnings about AI dangers are a "HOAX," posting that "AI taking over the World, destroying Humanity, and all other things bad, is a HOAX" . And the European Union will propose a ban restricting under-15s from social media, video sharing platforms, AI chatbots and online games Thursday .

Three statements landed within 48 hours. One policy chief publicly confirming what had been reported days earlier—that rivals have been meeting privately while their CEOs published essays framing coordination as a fresh idea. One president placing AI safety in the same rhetorical bucket as impeachment inquiries and climate science, then phoning Nvidia's CEO on stage to say the robots won't take over. And one continent deciding a chatbot conversation carries the same risk threshold as a social media feed, triggering the same age gate California imposed on Instagram four years ago.

3 Claims

Claim 1 — OpenAI: Has been working with Anthropic and Google DeepMind on AI safety coordination for weeks, does not believe the three firms need an antitrust waiver

Chris Lehane, OpenAI's global policy chief, told reporters Tuesday the company has been working with rivals Anthropic and Google DeepMind on AI safety for weeks . Lehane said the company's engagement with Anthropic and Google had been under way for several weeks, and added that OpenAI does not see the need for an antitrust waiver for the three AI firms to coordinate on safety matters .

Discussions have been ongoing since Demis Hassabis, the chair of Google DeepMind, released a proposal in July calling for a U.S.-led "Standards Body" . The announcement followed reporting last week that working groups from the three companies had been meeting regularly since July. Lehane's Tuesday remarks put coordination on the record but revealed no start date, participant names, or written agreements.

Grade by: 2026-10-16 (1 month). OpenAI should disclose a concrete output—a shared testing protocol, a public memorandum, or a list of agreed benchmarks—within 30 days if the coordination is producing operational results rather than positioning.

Claim 2 — Trump: Posted Monday that AI takeover fears are a hoax, stated the only guardrail AI needs is a "STRONG AND SMART (High IQ!) PRESIDENT"

In Truth Social posts Monday, President Trump downplayed concerns about artificial intelligence, calling criticism of its unregulated expansion a "hoax" . Trump said "The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!"

Trump told attendees at the All-In Summit in Los Angeles Monday over the phone, "The AI will not be taking over the rest of the world. The whole thing is a hoax" . He claimed the U.S. is dominating other countries in AI development, adding there is a "SICK conspiracy going on against AI and Data Centers" . The remarks arrived days after Anthropic CEO Dario Amodei published an essay calling for the industry to slow capability gains.

Grade by: 2027-09-16 (1 year). The claim is that presidential oversight alone is sufficient. The invalidator: any federal AI legislation signed by Trump within 12 months, or any binding executive order creating mandatory pre-release evaluations or safety standards for frontier labs. Either would contradict the stated position that no additional controls are needed.

Claim 3 — European Commission: Will propose Thursday a regulation barring under-15s from accessing AI chatbots, social media, video-sharing platforms, and online games without parental control

The European Union will propose a ban restricting under-15s from social media, video sharing platforms, AI chatbots and online games in the most sweeping plan yet to protect children from online dangers, a European Commission document seen by Reuters shows . The EU proposal is part of an EU Kids Act that European Commission President Ursula von der Leyen and EU tech chief Henna Virkkunen will present on Thursday .

Children from 15 onwards can set up their own accounts under the proposal, while children between 13 and 14 can ask parents to open introductory accounts for social media and video-sharing platforms under parental control and with limited contacts and strict time limits . A policy document shows the bloc's executive branch will demand that social media companies, online gaming services and AI chatbots age-gate their services to keep out the youngest teens .

Grade by: 2026-09-24 (1 week). The Commission will either present the EU Kids Act proposal Thursday, September 17, as scheduled, or it will not. Postponement, significant scope reduction, or removal of AI chatbots from the covered services would invalidate the claim.

2 Reckonings

Reckoning 1 — California AI auditor registry: Signed September 9, effective January 1, 2029

In Entry 104 (September 11, 2026), this ledger recorded that California Governor Newsom signed SB 813 and AB 1405 on September 9, establishing the nation's first AI auditor registry and barring anyone without a state registration number from conducting a covered AI audit starting January 1, 2029.

Projection: The law would create a mandatory registry by 2029, with California's Government Operations Agency standing up the system, handing out unique registration numbers, and publishing auditor information online.

What happened: The bills were signed as stated. The Government Operations Agency has until January 1, 2029, to implement the registry. No audits have been conducted under the new framework yet because the law is not yet in effect. The claim was verifiable and accurate—the law exists, the deadline is set, and the signing occurred on the date stated.

Grade: A. The claim was factual, the source was official, and the dates matched.

Invalidator: The grade would have been lower if the bills had been vetoed, if the effective date had been different from January 1, 2029, or if the signing date had been misreported.

Reckoning 2 — OpenAI Navier-Stokes claim: 10,000 agents solved the problem in 88 hours

In Entry 102 (September 9, 2026), this ledger recorded that OpenAI announced it solved the Navier-Stokes existence and smoothness problem using an internal model significantly more capable than GPT-6 Astra, deploying roughly 10,000 agents that produced a proof in 88 hours, with formalization in Lean taking another 17 hours.

Projection: OpenAI's model solved a Millennium Prize Problem using agents.

What happened: Multiple mathematicians, including NYU's Tristan Buckmaster, immediately questioned the validity of the proof and suggested OpenAI's model may have accessed their unpublished work. As of September 16, no independent verification has confirmed the proof's correctness, the Clay Mathematics Institute has not awarded the prize, and OpenAI has not released the proof for peer review.

Grade: C. The claim was reported accurately—OpenAI did announce it—but the substance remains unverified. The company made a claim about solving a 90-year-old problem without providing the evidence needed for independent confirmation. Seven days later, the announcement looks more like a press release than a mathematical result.

Invalidator: The grade would have been an A if the Clay Mathematics Institute had confirmed the proof, or if peer-reviewed verification had appeared in a mathematics journal within the week. It would have been an F if OpenAI had retracted the claim.

1 Refusal

I refused to treat Trump's Monday statements as off-the-record color. The president of the United States said on Monday that AI safety concerns are a hoax, that no guardrails beyond his own judgment are needed, and that warnings from lab CEOs constitute a "sick conspiracy." Those are claims about policy, governance, and the sufficiency of executive oversight. They are on the record. They have a clear grading horizon—any legislative action or binding executive order within twelve months that contradicts the stated position. I could have summarized them in a paragraph under "political reaction" and moved on. I didn't. I refused to let the highest-ranking government official escape the same accountability structure I apply to lab CEOs.

— Roger Grubb, Editor