The Transformational Artificial intelligence to Modernize the Economy against Extreme Weather and Wildfires Act (TAME) requires the Under Secretary at NOAA to develop curated training datasets, test AI‑based global and regional weather models, and explore AI tools for risk communication and wildfire readiness. It mandates technical assistance, model assessment frameworks, and avenues for public‑private co‑investment while preserving support for traditional numerical weather and observation programs.
The bill matters because it moves NOAA from experimentation toward operational adoption of machine‑learning weather models, mandates public release of models and associated federal data subject to IP and national‑security exceptions, and authorizes multi‑year funding — changes that reshape forecasting operations, vendor relationships, data access, and the agency’s workforce needs.
At a Glance
What It Does
The bill requires NOAA to develop and curate comprehensive weather‑forecasting training datasets within 4 years, test AI‑based global and regional weather models, and produce recurring reports on progress. It also directs NOAA to publish operational and experimental AI models and associated federal data at no cost, subject to national security and IP accommodations.
Who It Affects
NOAA and the National Weather Service; federal research partners (DOE, NASA, NSF), academic and private weather‑AI firms; emergency managers, fire response agencies, and state/local forecast offices that will test and adopt AI outputs; and commercial data providers whose contracts or IP may be implicated by public release requirements.
Why It Matters
By mandating curated training datasets, model evaluation frameworks, and open access, the bill accelerates integration of AI into operational forecasting and shifts where technical capability and value accumulate — from proprietary models to shared federal datasets and partnership arrangements — while building in high‑stakes carveouts for security and IP.
More articles like this one.
A weekly email with all the latest developments on this topic.
What This Bill Actually Does
The heart of the TAME Act makes the NOAA Under Secretary responsible for creating high‑quality, documented training datasets for AI weather models. The statute sets a four‑year clock for dataset development, requires NOAA to assess existing federal reanalysis products before building anew, and explicitly defines ‘‘curate’’ to include provenance metadata and periodic updates.
That means NOAA must move beyond ad hoc data pools to reproducible, versioned datasets suitable for machine‑learning training.
The bill authorizes NOAA to build and test both a global AI weather model and smaller‑scale regional or local AI models. It does not force replacement of numerical weather prediction (NWP); rather, it requires NOAA to continue funding observations, NWP research, data assimilation, and post‑processing.
At the same time, NOAA must explore hybrid approaches — ensembles that combine AI models with numerical models, uncertainty quantification research, and AI‑driven communication tools aimed at improving wildfire preparedness and impact‑based decision support.Operationalization receives parallel attention: NOAA must provide technical assistance and data access so forecasters, social scientists, emergency managers, and outside researchers can test AI outputs in testbeds. The bill pushes for a common assessment framework to compare AI and numerical models using historical backtests, and permits the Under Secretary to commission independent studies (for example, through the National Academy) about AI’s effects on the broader weather enterprise.On access and partnerships, the Act directs NOAA to publish operational and experimental AI models and the federal data that underlie them at no cost, subject to carveouts for national security, contract restrictions, trade secrets, and IP law.
It also authorizes NOAA to pursue co‑investment and novel partnership arrangements — including non‑federal contributions and shared IP rights — to accelerate high‑risk, high‑reward R&D. The Under Secretary may withhold models or data when national security requires it, and must deliver a classified/unclassified report within one year analyzing risks from foreign access to U.S. weather data.
The Five Things You Need to Know
The Under Secretary must develop and curate comprehensive weather‑forecasting training datasets within four years of enactment, including provenance metadata and periodic updates.
NOAA is authorized to develop and test an AI‑based global weather model and to experiment with regional and local AI models; the agency must still fund observations and numerical weather model research.
The Under Secretary must submit a report on subsection activities not later than 2 years after enactment and biennially through 2035 to Congressional science committees.
NOAA must make operational and experimental AI weather models and associated federal data publicly available at no cost, but may withhold releases to protect national security, IP, contract restrictions, trade secrets, or the agency’s life‑safety mission; a classified/unclassified risk report on foreign access is required within one year.
The bill’s authorization figures appear twice in the text: one version authorizes $311M for FY2026 and $76M annually 2027–2030, while an alternate inserted version lists $105M for FY2026 and $25M annually 2027–2030 — the text contains both sets of figures.
Section-by-Section Breakdown
Every bill we cover gets an analysis of its key sections.
Definitions tailored to AI and forecasting
This subsection sets precise definitions for ‘‘artificial intelligence,’’ ‘‘artificial intelligence weather model,’’ ‘‘curate,’’ ‘‘numerical weather model,’’ ‘‘observational data,’’ and ‘‘synthetic data.’nBy defining ‘‘curate’’ to require documented quality control and provenance metadata, the bill embeds data governance expectations into NOAA’s mandate — a practical signal that datasets must be auditable and continuously maintained, not just aggregated.
Training datasets and model development
NOAA must develop and curate comprehensive training datasets within 4 years, consulting multiple federal science agencies and advisory committees. The Under Secretary is instructed to evaluate existing federal reanalysis datasets to avoid unnecessary duplication. Practically, NOAA will need infrastructure for versioned dataset storage, provenance tracking, and access controls, and it must coordinate with partners who already host large observational archives.
AI models plus continued support for numerical methods
The statute authorizes building a global AI model and experimenting with regional/local AI systems while explicitly preserving funding and support for observational platforms, numerical models, data assimilation, and post‑processing. This creates an operational dual‑track: NOAA must scale AI R&D but keep existing numerical modeling pipelines and observing investments intact, which has implications for budget phasing and staff skills.
Advanced applications, uncertainty, and operational help
The bill pushes NOAA to research uncertainty quantification, ensemble generation from AI models, and AI tools to improve wildfire readiness and decision support. It requires technical assistance to forecasters and emergency managers and directs development of a common framework to assess AI and numerical models using historical comparisons. That framework will shape which models move from R&D into operational use and what verification metrics NOAA emphasizes.
Public‑private partnerships and data availability
NOAA may pursue novel co‑investment strategies with academia, industry, and international partners, including accepting non‑federal resources and sharing IP from R&D collaborations. The Under Secretary must create a plan to make models and federal data publicly available at no cost, but the plan includes express carveouts for national security, IP, trade secrets, contract restrictions, and life‑safety mission concerns. Implementing this will require new contract language, licensing reviews, and legal clearance processes for data redistribution.
Foreign‑access risk assessment and national‑security withholding
Within one year NOAA must deliver an unclassified and classified report analyzing economic and intellectual‑security risks from foreign access to U.S. weather data, and the Under Secretary can withhold models/data after consulting DoD where national security demands it. Agencies will need a risk‑classification routine to evaluate datasets and models for potential national‑security sensitivity and a clear decision pathway for redaction or restricted release.
Workforce recruitment and retention flexibilities
The Under Secretary may pursue novel hiring, retention, and exchange programs to compete with private sector salaries and skill demands. That authorization signals the need for HR policy changes, potential use of existing excepted‑service hiring authorities, and formalized staff exchange programs with academia and industry to retain AI and data engineering talent.
Authorizations (conflicting figures appear)
The text contains two distinct authorization schedules: an inserted version listing $311 million for FY2026 and $76 million annually for FY2027–2030, and an alternative set listing $105 million for FY2026 and $25 million annually thereafter. Either set would be the statutory funding ceiling; the presence of both figures in the bill text creates an ambiguity that would need resolution in final enrolled language or accompanying legislative action.
This bill is one of many.
Codify tracks hundreds of bills on Science across all five countries.
Explore Science in Codify Search →Who Benefits and Who Bears the Cost
Every bill creates winners and losers. Here's who stands to gain and who bears the cost.
Who Benefits
- Operational forecasters and emergency managers — They gain new tools, technical assistance, and model assessment frameworks to evaluate AI outputs, potentially improving lead times and impact‑based decisions (notably for wildfire response).
- Academic and research institutions — The requirement for curated public training datasets and independent assessments opens access to high‑quality data for reproducible research and accelerates collaboration on uncertainty quantification and hybrid modeling.
- Smaller weather‑AI innovators — Public availability of NOAA training datasets and experimental models lowers data barriers to entry, enabling startups and non‑profit groups to develop specialty products or local forecasting services without buying expensive observational archives.
Who Bears the Cost
- NOAA and the federal treasury — NOAA must stand up dataset curation infrastructure, manage public releases, conduct evaluations, and potentially hire and retain AI talent; the bill authorizes funding but leaves budgetary choices that affect other programs.
- Contracted commercial data providers — Providers whose datasets sit behind contracts or contain proprietary elements may face pressure to renegotiate terms or limit access if NOAA seeks broader redistribution rights tied to operational models.
- State and local forecast offices and emergency services — Adoption of AI outputs will impose training, integration, and operational‑testing burdens; these organizations will need to adapt procedures, evaluate AI reliability, and invest in staff capacity to use new decision tools.
Key Issues
The Core Tension
The central dilemma is between openness and control: publishing models and federal datasets accelerates scientific progress, lowers barriers for innovators, and supports reproducibility, but unfettered release risks revealing capabilities exploitable by hostile actors, violates third‑party IP or contract terms, and may expose operationally immature models to misuse — forcing NOAA to choose between transparency and safeguarding security, quality, and legal obligations.
The bill tries to thread three competing demands: rapid, open access to federal datasets and models; protection of national security and commercial IP; and safe, reliable operational forecasting. Translating ‘‘curate’’ into practice requires NOAA to build reproducible pipelines, metadata standards, and governance mechanisms — tasks that are labor‑ and compute‑intensive.
NOAA will need to manage provenance, licensing, and redistribution rights across legacy contracts and commercial partnerships, where some data may never be releasable without compensation or renegotiation.
Operationalizing AI models raises methodological and mission risks. Machine‑learning systems excel where training data are dense, but weather‑data coverage is uneven geographically and temporally; the Act encourages cost‑functions and synthetic data to mitigate this, but synthetic augmentation and domain shifts can introduce hard‑to‑diagnose biases.
Creating a common assessment framework is essential but technically fraught: backtests over historical periods may not capture future nonstationarity (climate change) and could understate edge‑case failures. Finally, the text contains conflicting authorization numbers, creating near‑term funding uncertainty that will shape how aggressively NOAA can hire personnel, procure compute, and execute co‑investment programs.
Try it yourself.
Ask a question in plain English, or pick a topic below. Results in seconds.