The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
$schema: string
title: string
description: string
type: string
items: struct<type: string, required: list<item: string>, properties: struct<country_code: struct<type: lis (... 4453 chars omitted)
child 0, type: string
child 1, required: list<item: string>
child 0, item: string
child 2, properties: struct<country_code: struct<type: list<item: string>, description: string>, program_name: struct<typ (... 4389 chars omitted)
child 0, country_code: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 1, program_name: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 2, program_status: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 3, timeline_start_year: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 4, timeline_target_year: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 5, announced_usd: struct
...
rmsTo: string
@context: struct<@vocab: string, cr: string, sc: string, data: struct<@id: string, @type: string>>
child 0, @vocab: string
child 1, cr: string
child 2, sc: string
child 3, data: struct<@id: string, @type: string>
child 0, @id: string
child 1, @type: string
version: string
recordSet: list<item: struct<@type: string, @id: string, name: string, description: string, field: list<item: s (... 171 chars omitted)
child 0, item: struct<@type: string, @id: string, name: string, description: string, field: list<item: struct<@type (... 159 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, field: list<item: struct<@type: string, @id: string, name: string, description: string, dataType: string, s (... 81 chars omitted)
child 0, item: struct<@type: string, @id: string, name: string, description: string, dataType: string, source: stru (... 69 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, dataType: string
child 5, source: struct<fileObject: struct<@id: string>, extract: struct<column: string>>
child 0, fileObject: struct<@id: string>
child 0, @id: string
child 1, extract: struct<column: string>
child 0, column: string
to
{'@context': {'@vocab': Value('string'), 'cr': Value('string'), 'sc': Value('string'), 'data': {'@id': Value('string'), '@type': Value('string')}}, '@type': Value('string'), 'conformsTo': Value('string'), 'name': Value('string'), 'description': Value('string'), 'url': Value('string'), 'license': Value('string'), 'creator': {'@type': Value('string'), 'name': Value('string'), 'url': Value('string')}, 'datePublished': Value('timestamp[s]'), 'version': Value('string'), 'identifier': Value('string'), 'keywords': List(Value('string')), 'dateModified': Value('timestamp[s]'), 'temporalCoverage': Value('string'), 'distribution': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'contentUrl': Value('string'), 'encodingFormat': Value('string'), 'sha256': Value('string')}), 'recordSet': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'field': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'dataType': Value('string'), 'source': {'fileObject': {'@id': Value('string')}, 'extract': {'column': Value('string')}}})})}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
$schema: string
title: string
description: string
type: string
items: struct<type: string, required: list<item: string>, properties: struct<country_code: struct<type: lis (... 4453 chars omitted)
child 0, type: string
child 1, required: list<item: string>
child 0, item: string
child 2, properties: struct<country_code: struct<type: list<item: string>, description: string>, program_name: struct<typ (... 4389 chars omitted)
child 0, country_code: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 1, program_name: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 2, program_status: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 3, timeline_start_year: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 4, timeline_target_year: struct<type: list<item: string>, description: string>
child 0, type: list<item: string>
child 0, item: string
child 1, description: string
child 5, announced_usd: struct
...
rmsTo: string
@context: struct<@vocab: string, cr: string, sc: string, data: struct<@id: string, @type: string>>
child 0, @vocab: string
child 1, cr: string
child 2, sc: string
child 3, data: struct<@id: string, @type: string>
child 0, @id: string
child 1, @type: string
version: string
recordSet: list<item: struct<@type: string, @id: string, name: string, description: string, field: list<item: s (... 171 chars omitted)
child 0, item: struct<@type: string, @id: string, name: string, description: string, field: list<item: struct<@type (... 159 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, field: list<item: struct<@type: string, @id: string, name: string, description: string, dataType: string, s (... 81 chars omitted)
child 0, item: struct<@type: string, @id: string, name: string, description: string, dataType: string, source: stru (... 69 chars omitted)
child 0, @type: string
child 1, @id: string
child 2, name: string
child 3, description: string
child 4, dataType: string
child 5, source: struct<fileObject: struct<@id: string>, extract: struct<column: string>>
child 0, fileObject: struct<@id: string>
child 0, @id: string
child 1, extract: struct<column: string>
child 0, column: string
to
{'@context': {'@vocab': Value('string'), 'cr': Value('string'), 'sc': Value('string'), 'data': {'@id': Value('string'), '@type': Value('string')}}, '@type': Value('string'), 'conformsTo': Value('string'), 'name': Value('string'), 'description': Value('string'), 'url': Value('string'), 'license': Value('string'), 'creator': {'@type': Value('string'), 'name': Value('string'), 'url': Value('string')}, 'datePublished': Value('timestamp[s]'), 'version': Value('string'), 'identifier': Value('string'), 'keywords': List(Value('string')), 'dateModified': Value('timestamp[s]'), 'temporalCoverage': Value('string'), 'distribution': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'contentUrl': Value('string'), 'encodingFormat': Value('string'), 'sha256': Value('string')}), 'recordSet': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'field': List({'@type': Value('string'), '@id': Value('string'), 'name': Value('string'), 'description': Value('string'), 'dataType': Value('string'), 'source': {'fileObject': {'@id': Value('string')}, 'extract': {'column': Value('string')}}})})}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Scrutica Sovereign AI Program Index
Government AI programs with separate amounts for announced funding, commitments and disbursements, and, where the government publishes it, the amount paid or spent against a budget, with its document and date. The dataset includes government-funded portions of announcements, qualitative assessments of dependence on NVIDIA and on US-jurisdiction inputs, and source notes on funding scope and disclosure.
Scrutica Sovereign AI Program Index — v2026.5
Release date: 2026-11-12
Documentation revised: 2026-11-12
License: CC-BY-4.0
Suggested citation: see CITATION block at the bottom.
Version history
- v2026.5, released 2026-11-12. Thirty columns beginning
paid_follow the thirty columns of v2026.4, which keep their names and order. They are filled for the 10 programmes whose government publishes what it has paid or spent against a budget: the amount in the currency of the document that states it, with that document, its page or table and its date; the budget the amount is paid against, with the document that states the budget; the share paid and the months of the budget’s period gone at the figure’s date; and the amount and the budget in US dollars, each with the rate it uses. In the thirty columns, France’sannounced_govt_only_usdandcommitted_usdare restated from the Cour des comptes report of 19 November 2025 (US$1.24B and US$977M, each US$2.72B in v2026.4), and France’s and India’sdisbursed_usdare filled from the governments’ own figures (US$340M and US$30.7M), with their conversions infunding_fx; the correction of 25 October 2026 gives the documents. France’s source list has 19 entries, from 17. The United States record’sprogram_namedrops the Stargate Project, a private joint venture, and Saudi Arabia’skey_partnersnames xAI as SpaceXSI, the name its terms of service use. - v2026.4, released 2026-09-25.
tsmc_dependencyandbis_jurisdiction_reachare empty on every row: neither rating had a published rubric or a cited source, and the TSMC rating read high on 35 of 36 programmes.nvidia_dependencyis empty for China, the Czech Republic, Poland and Ireland, whose records name no NVIDIA hardware and list no source about NVIDIA. Spain'sannounced_govt_only_usdis filled (EUR 1.5B at the record's own rate). - v2026.3, dated 2026-07-17. Its CSV was rewritten in place three times after release, so four different files were served under that one version: SHA-256 19a19e3f91fad40a2978b5b733f338eaa13f774f40906be003bf8133868dec56 (17 July, 34 programmes), 38c2dbda0a3acc41e571eab2fdd1a56c6b45faaa59dea982d3ae9ed230a81f13 (14 August, 36), 6be7b5ddc8b04ae01aaf36aabf9961045e051391a657f3488dcc9c5ca698a68f (14 August, 36) and cd296f9b045355c44b4696817e45fe61dd7af396c9f1c3dc456a2e4f8225304b (8 September, 36). A citation of v2026.3 identifies one of these only with its checksum. From v2026.4 the packager refuses to rewrite a released CSV.
Coverage
36 national and regional AI programme records. Funding is recorded separately as announced, committed and disbursed. announced_usd can include private investment in a program announcement; announced_govt_only_usd records government funding, which can cover a different period or set of initiatives. The paid_ columns are filled for the 10 programmes whose government publishes what it has paid or spent against a budget.
Funding-stage definitions
- Announced: funding publicly announced for the program, including private investment counted in that announcement.
- Committed: funding recorded as approved through government or corporate processes.
- Disbursed: reported funding transfers or payments. The source identifies the recipient and date; a transfer to an administering body can precede programme expenditure.
- Paid (the
paid_columns): what a government document reports paid out or spent against a budget (paid_measuregives its word, payments or expenditure), in that document’s currency, at the date it gives. Where the document prints the figure only in parts, each year or each sub-programme,paid_amountis their sum andpaid_is_sum_of_partsis true. The budget can be the whole programme’s, one component of it, or a wider budget that includes it (paid_scope), and its document can differ from the paid figure’s.
disbursed_usd is the programme record’s own field, filled only where that record converts a paid figure; the paid_ columns are read from the government spending record, which covers more programmes. Where both are filled (France and India), disbursed_usd equals paid_usd. The Sovereign AI Funding page’s gap to an even pace, in points, is 100 × (min(1, paid_months_elapsed ÷ paid_period_months) − paid_amount ÷ paid_budget_amount), rounded.
Amounts can refer to different dates and scopes. Some programs have missing stages or a commitment larger than the announcement field. Check the program disclosures before treating these amounts as successive portions of one budget.
A true *_is_estimated flag marks an estimated funding value, including source-published approximations and documented currency conversions. A null flag means the amount is absent or its estimate status has not been classified.
Sources and authority tiers
| Source | Authority tier | What it contributes |
|---|---|---|
| Scrutica sovereign-program compilation | 2 | Curated program records and their cited sources |
| Budget reports, accounts and audits of governments and public bodies | 1 | The paid_ columns: each paid figure and budget, with its document, locator and date |
The CSV records one cited URL per program, drawn from government and institutional disclosures, company announcements and published reporting. The source list can contain more entries than the URL exported here; its length is recorded in source_count. The compilation's tier does not assign an authority tier to each underlying document or funding value. Each paid figure and each budget in the paid_ columns carries its own document, address and tier.
Supplier dependence
nvidia_dependency: reliance on NVIDIA acceleratorsus_dependency: reliance on US-jurisdiction inputs, including chips, tools and intellectual propertytsmc_dependencyandbis_jurisdiction_reach: empty on every row from v2026.4 (see Version history). The columns stay so that files keep one layout across versions.
The rated fields permit none, low, medium or high. Null in JSON (blank in CSV) means the assessment is unavailable; none is a recorded classification. These are qualitative assessments of the programme’s chips, partners and supplier relationships. The file provides no numeric mapping for the categories.
Currency handling
Where recorded, currency_of_record gives the original announcement currency; fx_rate_used, fx_rate_date and fx_rate_source document the exchange rate used for USD conversion. Check the rate’s quoted units in its source. The funding_fx JSON column records conversions for individual funding fields; rates can differ between stages. Funding columns ending in _usd remain in US dollars.
The paid_ amounts stay in the currency their documents print. paid_usd converts paid_amount at paid_usd_rate, the reference rate of paid_usd_rate_date from paid_usd_rate_source, and paid_budget_usd converts the budget at a rate of the budget’s own date, each to three significant figures.
The packager copies the stored funding values without further currency or inflation adjustment.
Refresh path
scripts/package-datasets.ts packages the curated program records in data/processed/sovereign_programs.json. It copies the funding figures, expands the dependency assessments into separate columns, joins chip and partner names with semicolons, and counts the entries in each program’s source list. For each programme whose government publishes a paid figure, it reads the paid_ columns from the line of the government spending record (data/processed/sovereign_government_spending.json) that the Sovereign AI Funding page prints. It refuses to package a programme whose disbursed_usd and paid_usd differ.
The CSV has no dedicated columns for hardware quantities, partner allocation counts or their evidence; those details remain in the programme records. Counts may also appear within the exported notes.
Known limitations
- Funding stages have different levels of coverage. A blank amount is unknown in this file.
key_notesdescribes qualifications to the funding figures, including announcements that combine several initiatives or private investment.structural_non_disclosuremarks programs assessed as having limits on disclosure of material capacity or spending;structural_non_disclosure_reasongives the recorded explanation where available.- The budgets in the
paid_columns differ in scope and length (paid_scope,paid_period_months), so their shares paid do not rank programmes against each other.
Citation
@dataset{scrutica_sovereign_ai_index_2026_5,
author = {Gringras, David and {Scrutica}},
title = {Scrutica Sovereign AI Program Index},
year = {2026},
version = {2026.5},
publisher = {Scrutica},
note = {No DOI assigned; cite the version and SHA-256 checksum at /datasets/sovereign-ai-index},
url = {https://scrutica.com/datasets/sovereign-ai-index}
}
- Downloads last month
- 50