Mandate
Review methodology, conflicts, AI assurance and material incidents—not daily headlines or commercial sales.
PUBLIC STANDARD / VERSION 1.0
Trust is an operating system, not a claim.
This charter defines who can oversee GOV.TODAY Intelligence, how AI may be used, how evidence is handled and where commercial or policy outreach must stop.
OUR COMMITMENT
Independent findings. Named accountability. Visible limitations.
GOV.TODAY does not sell a predetermined conclusion. Clients may define a question and delivery window; they do not rewrite evidence, hide material limitations or prevent a necessary correction.
01 / INDEPENDENT OVERSIGHT
A seven-seat Advisory Board balanced by design.
Founding appointments are open. No individual is presented as appointed until consent, conflict review and public disclosure are complete.
7INDEPENDENT SEATS
Academics protect methodological depth. Practitioners test operational relevance. Independent ethics expertise protects the public-interest boundary.
01
Independent chair — methodology and data governance
02
Two public-administration or security-policy academics
03
Two former digital-government or regulatory practitioners
04
One OSINT, sanctions or trade-data specialist
05
One independent ethics or civil-society member
Independence
Two-year staggered terms, annual conflict disclosures and recusal from affected decisions.
Transparency
Public charter, membership, meeting summaries and an annual assurance statement.
02 / AI ASSURANCE
AI can accelerate analysis. It cannot own accountability.
The control model is aligned to NIST AI RMF, the OECD AI Principles and relevant EU AI Act obligations. Legal applicability is assessed use case by use case.
Human accountability
A named editor owns every high-impact conclusion and the final publication decision.
Provenance
Sources, retrieval dates, transformations and model-assisted steps are recorded.
Testing
Material outputs are checked for factual error, bias, leakage and unsupported inference.
Explainability
Decision-relevant claims must be understandable without access to a proprietary model.
Security
Access is role-based; sensitive work uses minimisation, retention limits and incident response.
Override
Editors can stop, correct, roll back or decommission any automated workflow.
03 / EDITORIAL CONTROL
A verifiable chain from source to conclusion.
Every consequential statement must be reproducible by a qualified reviewer from the recorded evidence and method.
Source tiers
Primary law, official records and datasets lead. Secondary reporting adds context and is identified as such.
Two-person review
High-risk allegations, named entities and consequential forecasts require an independent second reviewer.
Confidence labels
Confirmed fact, analytical assessment, scenario and unknown are never blended into one claim.
Corrections protocol
Material errors are corrected visibly, timestamped and propagated to affected client deliverables.
LAWFUL OSINT & DATA MINIMISATION
Only lawful sources are collected. Personal data is minimised and published only when a documented public-interest test is met. The evidence log distinguishes observation, attribution and analytical assessment; named subjects receive a meaningful right of reply when feasible.
04 / LEGAL & PROCUREMENT GATES
Outreach begins only after the representation question is answered.
Policy research, lobbying, representation for a foreign principal and government contracting are not interchangeable. Each engagement is classified before outreach, proposal or payment.
- U.S. FARA and LDA assessment for relevant activities and clients.
- EU Transparency Register assessment before covered interest representation.
- Sanctions, export-control, privacy, conflicts and anti-bribery screening.
- SAM.gov or other procurement registration only when the contracting route requires it.
05 / DISCLOSURE REGISTER
What GOV.TODAY will publish as the hub scales.
- 01
Advisory Board membership, bios, terms and conflicts.
- 02
Funding categories and material institutional sponsors.
- 03
Method versions, corrections and significant AI incidents.
- 04
Sponsored-research labels and editorial-firewall statements.
IMPLEMENTATION
Open the launch blueprint