Sunday, September 13, 2026HotTea archive editionVerified 2:55 AM PDT

The lead

Anthropic put permanent outside evaluators inside the company

Dario Amodei says third-party reviewers will get employee-level access as Anthropic calls for slower frontier development.

Listen to this edition

Prefer audio? The briefing has chapters and a full transcript.

The briefing

The rest of the morning

4 more stories

02

Sam Altman killed OpenAI's 2026 IPO as the safety alarm spread

OpenAI chief executive Sam Altman told Fortune that the company will not go public in 2026. He called the current moment ill-advised for an offering because of unfinished safety and alignment work. He said OpenAI will wait until both the business and society are ready.

An interview can move a timetable without creating a binding commitment. OpenAI has not published the proposed pact, its participants, its triggers or its enforcement. Altman did not identify a training run the company has paused under a shared rule. The IPO decision also follows earlier reporting about market volatility and OpenAI's finances, so safety is not the only possible pressure on timing.

Fortune ↗TechCrunch ↗The Verge ↗
Read article →
03

Coding jobs took only 28% of UK computer-science graduates

The share of recent UK computer-science graduates entering coding or programming jobs fell from about 40 percent previously to 28 percent last year. The analysis was prepared for the Guardian University Guide. The share entering any graduate-level occupation fell from more than 60 percent two years earlier to 50 percent.

The employment shift is observed. Its cause is not. HESA's survey does not show that employers replaced specific entry-level jobs with AI. Charlie Ball of Jisc said the software labor market clearly changed but remained reluctant to assign the change to AI without stronger evidence. A weak hiring cycle, employer budgets and graduates moving into cybersecurity or network roles can also affect the totals.

The Guardian ↗Higher Education Statistics Agency ↗
Read article →
04

Former EPA officials counted 30 federal moves that put AI growth ahead of pollution controls

The Environmental Protection Network is a group of former EPA employees. It identified 30 federal actions since January 2025 that it says weaken pollution safeguards, public review, science or enforcement as data centers expand. Seventeen of the actions explicitly cite AI or data centers.

The 30-action count and policy judgment come from an advocacy group of former EPA officials, not a current agency finding. The health figures are scenario estimates from a separate study, not observed deaths or medical bills. The authors say their model does not measure the additional effect of the 30 actions. EPA argues that its planned rules can reduce delays while retaining environmental protection.

Environmental Protection Network ↗The Verge ↗U.S. Environmental Protection Agency ↗
Read article →
05

Deepfake ads are turning creators into unpaid salespeople

The Guardian documented several cases in which generated images or videos used creators' likenesses to promote products they did not endorse. Lifestyle creator Emily Schuman found versions of herself in ads or posts for a telehealth company, a makeup brand and a blood-testing company. Gala and Meroda did not respond to the Guardian. Superpower said the post using Schuman was made by a scammer and that the company reported it.

The named cases show a real enforcement problem, but they do not establish how common creator impersonation is across all ads. The Tranchin allegations have not been decided by a court. Meta's removal and account totals are company figures, and broad loss estimates for deepfake scams use methods that are difficult to audit. The story supports the existence of harm, not a precise global prevalence rate.

The Guardian ↗Bloomberg Law ↗Meta ↗
Read article →

Analysis

A slowdown promise becomes real only when someone outside the company can test it

Anthropic offered access, and OpenAI offered time. Both moves can be measured. Neither yet gives an outside reviewer authority over a release.

1

Access comes before assurance

An evaluator cannot verify an internal safety process from a launch post. Standing access to models, logs and test results creates a chance to identify a problem before a release decision is final.

2

A delayed IPO is not a delayed model

OpenAI removed one capital-market deadline. It did not announce a model pause. The public should keep those decisions separate until a written safety pact defines what actually slows down.

3

Causation still needs work

The labor, pollution and deepfake stories each show visible pressure. They also show why observed outcomes, modeled estimates and causal claims should not be treated as the same proof.

The watchlist

Signals that could change the read

Anthropic names the outside evaluator and publishes its model, log, reporting and release-review rights.
OpenAI and other labs publish participants, capability triggers, verification rules and consequences for noncompliance.
The next UK cohort and employer records show whether junior software roles keep falling and which tasks changed.
EPA or state regulators publish project-level permits, monitoring data and modeled health effects for new AI loads.
Across the desks The brakes need receipts

Anthropic and OpenAI now say safety can justify slower development and delayed capital plans. The test is whether outside reviewers, public rules and measurable consequences follow the words.

Evaluator accessPermanentAnthropic says outside evaluators will receive company badges, desks, laptops and access close to internal risk teams.
OpenAI IPONot in 2026Sam Altman said a public offering this year would be ill-advised while safety work remains unfinished.
Coding-job share28%The share of UK computer-science graduates entering coding or programming jobs fell from about 40 percent previously.
Federal actions30 countedFormer EPA officials listed 30 federal moves they say weaken pollution protections as data-center construction expands.

Editorial direction, not a financial index. Each signal is tied to this edition’s reporting.

Edition validated · 5 stories · 14 unique sources

About HotTea & our sources →