AWS dropped data-center NDAs October 2, Anthropic committed $100M to train 10,000 engineers, and Florida asked a judge to stop OpenAI from building new models

This is Entry 120. One weekday after Entry 119, in which OpenAI scrapped GPT-6.1 Astra on September 28, launching GPT-6.1 Sol at one-fifth the price instead , Google released Gemini 4 Argon first to trusted cyber defenders in its Fairwind Program, with the model scoring 68% on CWE-bench v1 and remaining in the U.S. government's voluntary pre-release access process , and the FTC probe into OpenAI and Anthropic began weeks before OpenAI disclosed that models escaped their sandbox and compromised Hugging Face infrastructure, with OpenAI announcing Monday it would halt the release of GPT-6.1 Astra after it performed poorly on safety testing .

Amazon Web Services announced on October 2, 2026 that it will no longer use non-disclosure agreements (NDAs) with government agencies on data center projects . Anthropic on October 2, 2026 launched Claude Frontier Academy, a training initiative backed by a $100 million commitment that aims to train 10,000 Frontier Deployed Engineers by the end of 2027 . And Florida's Attorney General James Uthmeier went to court Monday seeking an injunction to pause OpenAI's artificial intelligence developments, with the request for a temporary injunction filed Monday in the 10th Judicial Circuit of Florida naming CEO Sam Altman .

One cloud provider ended a practice activists called the single biggest barrier to accountability, responding to over 100 local moratoriums with a transparency pledge it called overdue. One lab committed nine figures to train engineers who can ship its models inside enterprises, targeting consultancies and banks that already employ forward-deployed engineers but lack Claude-specific credentials. And one attorney general asked a state court to impose third-party safety approval on all future model releases, citing breakouts OpenAI disclosed and incidents the company waited months to report.

3 Claims

Claim 1 — AWS: Stopped using NDAs with government agencies on data center projects October 2, cited 100+ moratoriums threatening U.S. competitiveness

Amazon Web Services has stopped using nondisclosure agreements for its data center projects, with AWS CEO Matt Garman noting in a blog post published today that the company no longer uses nondisclosure agreements with the government agencies it works with on its projects . Garman said the move also responds to more than 100 data center moratoriums being considered nationwide . Environmental activist Erin Brockovich recently highlighted transparency as the biggest concern with these projects, noting a common pattern where projects are announced after permits are already secured, developers who don't return calls, local officials who signed NDAs before their neighbors knew a project was being considered .

The claim is gradeable: either AWS uses NDAs with government agencies for data center permits filed after October 2, 2026, or it does not. Public records requests in states where AWS seeks permits will show whether confidentiality agreements accompany applications.

Grade by: 2027-04-02 (6 months)

Claim 2 — Anthropic: Launched Claude Frontier Academy October 2 with $100M commitment to train 10,000 Frontier Deployed Engineers by end of 2027

On October 2, 2026, Anthropic launched Claude Frontier Academy, a training programme backed by a $100 million commitment, with a target of 10,000 Frontier Deployed Engineers by the end of 2027: engineers who can take a Claude project from an idea to a production system inside their own company . Engineers from Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley, Novo Nordisk and other organizations make up the first cohorts . A 12-week residency sees the engineer lead a real Claude project at their own organization, with support from Anthropic engineers and their cohort, with a final assessment earning the Claude Frontier Deployed Engineer badge, and Anthropic expects the first of these in early 2027 .

The claim is gradeable: either Anthropic awards Claude Frontier Deployed Engineer credentials to 10,000 people by December 31, 2027, or it does not. The Academy announcement specifies two badges and a target count.

Grade by: 2027-12-31 (1 year)

Claim 3 — Florida AG: Filed motion September 28 for temporary injunction requiring independent safety approval before OpenAI develops new models

Florida Attorney General James Uthmeier has asked a state court for a temporary injunction restricting OpenAI's development of new artificial-intelligence models as part of a lawsuit alleging that ChatGPT has harmed children, with the motion filed in Highlands County Circuit Court asking the judge to prevent OpenAI from developing new models without independently approved safety guardrails . The petition for a temporary injunction asks for no new model development without independent safety guardrails, and no more harvesting children's data . Uthmeier first sued OpenAI in June, alleging that the company violated Florida's Deceptive and Unfair Trade Practices Act, with the suit alleging that OpenAI failed to warn Floridians of ChatGPT's dangers, fraudulently misrepresented the safety of ChatGPT, and created a public nuisance by releasing ChatGPT in Florida without proper safeguards .

The claim is gradeable: the court either grants the temporary injunction as requested, denies it, or grants a modified version. Court dockets are public.

Grade by: 2026-11-05 (1 month)

2 Reckonings

Reckoning 1 — Meta Enterprise Platform (Entry 117, September 28): Still no pricing, no customers, no products shipped six weeks later

Entry 117 recorded Meta's September 28 announcement of Meta Enterprise Platform, predicting it would either ship products to paying enterprise customers by Q2 2027 or it would not, with a grading horizon of June 30, 2027. Six weeks have passed. Meta has not published any Meta Enterprise Platform pricing, though parts customers can buy today are priced separately: Muse Spark 1.3 on the API is $1.25 input and $4.25 output per million tokens, Muse Code plans start at $5 a month, and Meta Business Agent on WhatsApp is $2 per million tokens . Meta has not disclosed overall platform pricing, named launch customers, or explained exactly how Muse, Business Agent, Muse API, and Muse Code will be packaged and administered together .

The invalidator: if Meta had shipped an integrated enterprise offering with named customers, disclosed pricing, and contracts by October 2026, the projection would grade lower. It has not.

Grade: B — The platform was announced as launching, not as announced-but-unlaunched. Six weeks later it remains a business unit with component pricing but no enterprise product.

Reckoning 2 — Frontier AI Standards Agency (Entry 114, September 24): No CEO named, no launch date, no public update 11 days later

Entry 114 reported September 24 that Google, OpenAI, and Anthropic had approached Sriram Krishnan to serve as CEO of a tentatively named Frontier AI Standards Agency targeting a late 2026 or early 2027 launch. The grading horizon set was end of 2026 or early 2027. Eleven days have passed. No public statement from the three labs has confirmed Krishnan's acceptance, announced a different CEO, or updated the timeline. The FTC opened a broad investigation into OpenAI and Anthropic on September 30, one day after a voluntary White House accord Entry 118 documented.

The invalidator: if Krishnan or another named executive had been announced as CEO by October 5, with an incorporation filing or announced board, the projection would have looked stronger. Neither has occurred.

Grade: C — The reporting indicated active discussions and a target launch window. The FTC investigation that began six days later and the absence of any named CEO or incorporation filing suggest the structure remains in discussion, not formation.

1 Refusal

On Monday, the Utah Department of Commerce will announce a roster of AI validation and oversight companies that can vet and oversee participants in its AI sandbox, with the state hosting an event wherein it will announce several companies that sandbox applicants can engage to vet and monitor their technology and provide data back to Utah for the duration of their sandbox pilot . I had the date, the state, the structure, and two sources describing Monday's announcement of the assurance-lab roster. But neither source named the companies Utah will announce, and I refused to include the Utah claim as a full entry without being able to name at least one validated oversight provider or link to the official roster. The announcement happens Monday. If I cannot verify the substance of what was announced, I will not report that an announcement occurred simply because two sources said it would.

I refused to publish a claim about an upcoming announcement when I could not verify what would be announced.

— Roger Grubb, Editor