THE SIGNAL IN ONE SENTENCE
Open source is often described as an exit door. That is true in the same way that owning a map means you can cross a mountain. The map matters. So do boots, weather, food, a route that is not blocked and people who know where they are going. Kenya has now made open systems a more explicit part of its national digital plan. The Ministry of Information, Communications and the Digital Economy announced on November 17 that Kenya had joined the Digital Public Goods Alliance and the 50-in-5 campaign. The alliance welcomed Kenya as its newest member at its annual meeting in Paris the following day. The announcement is broader than a promise to publish code. The alliance says Kenya plans to operationalise its National Open Source Programme Office as a government adoption hub, expand training in open-source development, security and governance from 70 to more than 200 civil servants, deploy the Kenya Sovereign Open Source Digital Stack, known as KenStack, at the Konza National Data Centre and nominate at least two Kenyan open-source solutions for assessment against the Digital Public Goods Standard. Kenya also joined 50-in-5, a country-led campaign that aims to help 50 countries design, implement and scale at least one component of digital public infrastructure by the end of 2028. The campaign describes those components as secure and interoperable networks that can include digital identity, payments and data exchange. Those are meaningful commitments. They are not yet evidence that a public system can leave its current supplier, cloud, database, identity service or operating team without hurting the people who depend on it. The plain signal is this: test the exit before calling the system sovereign. Run a real service through a controlled departure. Export the data, configuration, audit history, access rules and documentation. Rebuild the service in a clean environment. Replace one important dependency. Restore from backup. Process a complete citizen journey. Measure the time, cost, missing knowledge, security exposure and service interruption. Then publish the result. This is not an argument against Kenya's move. It is the practical test that can make the move useful. The Digital Public Goods Standard already points toward several ingredients. Its nine indicators include open licensing, clear ownership, platform independence, documentation, a mechanism for extracting data, privacy, applicable law, security and other best practices. The standard also includes an important disclaimer: recognition focuses on the design of the core solution. The alliance does not evaluate local implementations as part of the review. That distinction is the whole story. A core product can be open while a national implementation is hard to move. The code may be public, but the deployment scripts may belong to a contractor. The database may be exportable, but the relationships, permissions and audit history may not survive the move. The software may run on several platforms in theory, but only one team may know how the production system actually works. Open code can reduce one kind of lock-in while contracts, operations, skills and infrastructure quietly rebuild it somewhere else. So begin with an inventory. Name the public service, the responsible agency and the people who rely on it. List every component needed to complete a real transaction: identity checks, payment rails, messaging, document storage, data exchange, analytics, fraud controls, support tools, security monitoring, hosting, backups and recovery. For each component, name the owner, maintainer, license, supplier, hosting location, data format, interface, contract end date and replacement option. Do not hide the small dependencies. A service can be theoretically portable and still fail because a one-time-password provider cannot be changed, a certificate is controlled by a contractor or an undocumented batch job produces the daily reconciliation file. Then define the exit event. It should be specific enough to fail. Move one non-trivial service from its current environment into a second approved environment. Switch one external provider. Rebuild from the published source and documented configuration. Restore a recent backup. Rotate the credentials. Run the system long enough to reveal the boring operational gaps. The second environment does not need to become permanent. This is a fire drill for dependence. Preserve a start receipt before anything moves. Record the software commit, container images, infrastructure configuration, database schema, model version where AI is involved, access roles, cryptographic keys under controlled custody, unresolved incidents and current service metrics. That receipt gives the exit test a baseline. Next, export the information. Data portability is not a download button that produces a giant file no one has opened. Another qualified team should be able to import the records, preserve their relationships, understand their meaning and reconstruct the decisions attached to them. Export the data dictionary, identifiers, timestamps, consent or authority records, retention rules, corrections, appeals, payment references and audit events. Preserve provenance so a transaction does not become an orphaned row after the move. Test the export before the contract is in trouble. Now rebuild. Give a Kenyan public-sector team the source, documentation and approved secrets. Ask them to create the service in a clean environment without the original vendor quietly doing the difficult parts off camera. Record every question they must ask. An undocumented question is a dependency. If the team cannot identify a compatible database, configure monitoring, renew a certificate, reproduce the deployment or explain a security control, the exit is not ready. The answer is not to shame the team. It is to make the missing knowledge visible and fund its transfer. This is where Kenya's civil-service training commitment can become more than a participation count. Do not report only that more than 200 people attended open-source training. Give a trained team a service and ask it to rebuild, patch, restore and operate the system. Measure what they can do independently, what still requires external help and how long each handoff takes. Skills are demonstrated in maintenance, not attendance. Then switch the dependencies. Replace a cloud service, identity connector, messaging provider or payment adapter using the documented interface. If the component cannot be replaced, say why. If replacement requires a long outage, proprietary conversion or supplier permission, record that as lock-in. Interoperability should be observed at the boundary. It is not enough for two systems to claim the same standard. A citizen record, payment status or service request should move across the boundary with the same meaning, permissions and appeal rights. For AI-enabled public services, add another layer to the test. Kenya's membership covers open AI systems as one category of digital public goods, but this announcement is a digital infrastructure programme, not proof of an AI deployment. The relevance is what happens when models become part of those shared public rails. Record the model, training or reference data, evaluation set, prompt or policy configuration, decision threshold and human-review boundary. Test whether a new model can replace the old one without erasing case history or changing a citizen's rights. Preserve the output and evidence needed to challenge an automated recommendation. Open weights alone do not make an AI service portable. The serving stack, feature pipeline, safety controls, hardware requirements, language data, monitoring and appeal process must travel too. Keep the public service running during the drill. This is where exit planning stops being an IT exercise. Pick citizen journeys that include success and trouble: a new application, an update to a record, a payment, a refund, a failed identity check, an offline or low-bandwidth attempt, an accessibility need, a correction and an appeal. Run them before and after the move. Check whether receipts, reference numbers, status messages and case histories remain intact. Check Kiswahili and other supported language paths. Check whether a person can still reach help when the new environment is partially degraded. The citizen should not become the integration test. Build the rollback path at the same time. An exit drill may expose a serious gap. The team needs a clear stop condition, a safe way to return to the previous environment and a record of every transaction created during the test. Rollback should preserve new data and avoid duplicate payments, lost cases or split histories. Security moves too. Rotate keys and privileged credentials. Verify who controls the domains, certificates, signing keys and software repositories. Recreate monitoring and alerting. Patch a dependency in the second environment. Restore from a backup that has not been prepared specially for the drill. An open repository that no local team can patch during an incident is not operational sovereignty. Procurement must support the exercise. Contracts should require current documentation, source access, export formats, interface specifications, configuration records, incident history, training, transition assistance, deletion evidence and a maximum time for handover. They should state which artifacts belong to the public authority and which restrictions survive termination. Do not wait for the final month of a contract to discover that the exit clause describes a meeting rather than a deliverable. Measure the test honestly. Record the time needed to freeze the baseline, export data, rebuild the service, replace dependencies, restore operations and close the security gaps. Record staff hours, outside assistance, failed steps, inaccessible artifacts, service interruption and the citizen journeys that did not survive intact. Publish the score by layer. Code may be portable while data is not. Data may travel while operations remain vendor-dependent. Operations may transfer while a critical payment or identity interface does not. A single green badge would hide the useful part. The public result should name the gap, owner, remedy, deadline and retest date. Some details should remain controlled for security and privacy. Public accountability does not require publishing credentials, exploitable configuration or personal data. It does require disclosing the boundary, the independent reviewer and enough evidence to show that the test happened. Kenya's new commitments offer a good place to start. Use KenStack deployment at Konza as a chance to publish an architecture and dependency register. Use the National Open Source Programme Office as the owner of repeatable exit drills and local maintenance evidence. Use the two planned DPG nominations to show both core-product conformance and local implementation readiness. Use civil-service training to build teams that can perform the drill, not just describe it. Then share the failure report through 50-in-5. Countries learn more from the undocumented connector, the failed restore and the missing contract clause than from another conference slide about openness. The announcements do not publish a budget, delivery calendar, technical inventory, completed exit test or measured service outcome. The alliance's more detailed note names planned activities, not finished results. That is fine. This is the beginning of a programme. But the standard for success should be set now. Kenya has said open source is an economic and sovereignty choice. Sovereignty is not the absence of partners. Public systems will always depend on specialists, vendors, communities and shared infrastructure. Sovereignty is the ability to understand those dependencies, choose among them, change them and keep public service intact when the relationship changes. The repository is the map. The exit drill proves the road exists.
01
WHAT ACTUALLY CHANGED
Kenya announced on November 17 that it had joined the Digital Public Goods Alliance and the 50-in-5 campaign.
The alliance welcomed Kenya as its newest member at its annual meeting in Paris on November 18.
Kenya plans to operationalise its National Open Source Programme Office as a government adoption hub and regional reference point.
The alliance says Kenya will expand training in open-source development, security and governance from 70 to more than 200 civil servants.
Kenya plans to deploy the Kenya Sovereign Open Source Digital Stack, known as KenStack, at the Konza National Data Centre.
At least two Kenyan open-source solutions are planned for assessment against the Digital Public Goods Standard.
The 50-in-5 campaign aims to help 50 countries design, implement and scale at least one digital public infrastructure component by the end of 2028.
The DPG Standard includes open licensing, clear ownership, platform independence, documentation and a mechanism for extracting data among its indicators.
The alliance states that DPG recognition evaluates the core solution and does not evaluate local implementations.
The announcements do not publish a budget, delivery calendar, technical inventory, completed exit test or measured public-service outcome.
02
WHY THIS MATTERS
Open source can reduce code dependence while contracts, skills, hosting, data formats and operational knowledge preserve vendor lock-in.
Digital identity, payments and data exchange are shared public rails, so a failed transition can interrupt many services at once.
A controlled exit drill reveals dependencies while the current supplier and operating team are still available to repair them.
Data export is useful only when another qualified system can preserve meaning, permissions, history and appeal records.
Local capacity is demonstrated when a team can rebuild, patch, restore and operate a service, not when people merely complete training.
AI components add model, data, serving, monitoring and human-review dependencies that must survive replacement.
Testing real citizen journeys protects people from becoming the first users to discover that a supposedly portable service does not work.
Publishing failures, remedies and retest dates can make shared international learning more practical than a generic success story.
03
WHERE IT COULD HELP
- Choose one non-trivial public service and define a controlled exit event with an accountable owner.
- Inventory every identity, payment, messaging, storage, data-exchange, hosting, security and support dependency.
- Record owners, maintainers, licenses, interfaces, data formats, contract dates and replacement options for each component.
- Freeze a baseline receipt containing code commits, images, configuration, schemas, access roles, incidents and service metrics.
- Export records with identifiers, relationships, metadata, permissions, corrections, appeals and audit history intact.
- Ask a Kenyan public-sector team to rebuild the service in a clean environment using the available documentation.
- Log every undocumented question and convert it into a knowledge-transfer requirement.
- Switch at least one cloud, identity, payment or messaging dependency through a documented interface.
- Restore a recent ordinary backup and verify that the recovered service passes security and transaction checks.
- Rotate privileged credentials, certificates and signing keys under public-authority control.
- Run complete citizen journeys for applications, payments, failures, corrections, accessibility and appeals.
- For AI components, preserve model versions, evaluations, decision thresholds, evidence and human-review boundaries.
- Test rollback without losing new transactions or creating duplicate payments and split histories.
- Measure time, cost, staff effort, outside help, service interruption, failed steps and unresolved dependencies.
- Publish layer-by-layer results with a gap owner, remedy, deadline and retest date.
KEEP A HAND ON THE WHEEL
Kenya's membership in the Digital Public Goods Alliance and 50-in-5 is a programme commitment, not proof that a particular public service is portable or free of vendor dependence. The alliance names planned activities including civil-service training, KenStack deployment and two DPG nominations. It does not publish completed deployment results, a technical architecture, procurement terms, a budget, a delivery calendar, an independent security assessment or a service-level outcome. The Digital Public Goods Standard assesses the core solution. The alliance explicitly says it does not evaluate local implementations as part of DPG recognition. A product's DPG status therefore should not be presented as certification that a Kenyan deployment is secure, inclusive, interoperable or easy to exit. The 50-in-5 campaign is an advocacy and learning campaign. Participation does not guarantee that a country has completed or successfully scaled a DPI component. Exit testing must protect personal data, credentials and security-sensitive configuration. Public reporting should disclose the test boundary, method, result, reviewer and unresolved gaps without publishing exploitable secrets or citizen records. Watch for Kenya's DPI Roadmap, KenStack architecture, named homegrown DPG nominations, repository and license details, procurement and transition terms, training assessments, operational ownership, citizen testing, accessibility evidence, security reviews, data-export demonstrations, completed exit drills and published retest results.
04
TERMS WORTH KEEPING
OPEN GLOSSARY CARD
Exit plan
A tested process for ending or changing an AI service while preserving records, revoking access and keeping essential work running.
OPEN GLOSSARY CARD
Vendor lock-in
Dependence on one supplier that makes switching costly or impractical because data, interfaces, skills, contracts or essential components cannot move easily.
OPEN GLOSSARY CARD
Digital sovereignty
The practical ability to control important digital systems, data, rules, suppliers, and continuity rather than depending on one outside service.
SOURCES AND VERIFICATION STATUS
This article was written from the materials below. Product claims and dates were checked against those sources on November 19, 2026.
THE PUBLICATION ENGINE
WANT A SIGNAL OF YOUR OWN?
We build source-grounded publications, private briefings, and editorial systems for organizations with something useful to say.
WORK WITH US