Newsroom

What evidence Kenyan election courts actually accept

8 min read

What this is / is not. This piece discusses general patterns in election integrity infrastructure and the public legal framework. It does not endorse or critique any specific candidate, party, or past electoral outcome, and it is not legal advice.

The framework, briefly

Election petitions in Kenya are governed primarily by Article 87 of the Constitution and the Elections Act, 2011, with operational rules issued by the Independent Electoral and Boundaries Commission. The petitioner must establish, on evidence, that an irregularity occurred and that it was of a magnitude that affected the result. The two halves of that sentence carry different evidential burdens. The first is about whether a thing happened. The second is about how much it mattered.

Both halves are won or lost on the quality of the underlying record. The platforms a campaign or observer uses during the election determine what that record looks like by the time it reaches a court. By then the choices that matter have already been made.

What admissibility actually requires

Three properties separate evidence a court treats seriously from material a court sets aside.

Chain of custody. A document or data point is only as strong as the path it took from origin to filing. A photograph of a Form 34A is meaningful if the record shows which agent took the photograph, at which station, at what time, who reviewed it, and what was changed between capture and export. A photograph that arrives as a JPEG attached to an email, with no provenance, is weak material no matter how accurate it is. The chain is the substantive claim.

Trusted timestamps. When a contested record was created, observed, or modified is often more important than what it says. A timestamp the system itself assigned - which the system can also rewrite - is not the same as a timestamp issued by an independent time-stamping authority under RFC 3161 or its IEC equivalents. Trusted timestamps make statements like "this photograph existed before the announced constituency total was published" possible to verify. Without them, sequence has to be reconstructed from context, and opposing counsel will reconstruct it differently.

Signed, immutable exports. A spreadsheet that a witness emails to a lawyer can be edited, accidentally or otherwise, between transmission and filing. A signed export with a cryptographic hash and an audit log showing every version is something a court can treat as a single document rather than a moving target. The standard tools here are ordinary: SHA-256 content hashes, RFC 3161 timestamps, append-only logs that record reads as well as writes. None of this is exotic. All of it has to be designed in from the start, because retroactive chain of custody is a contradiction in terms.

The "magnitude" half

Establishing that an irregularity occurred is necessary but not sufficient. The petitioner also has to show the irregularity was material to the result. This is where station-level granularity does load-bearing work. A claim that a county total is wrong is hard to size without examining the rows that compose it. A claim that 320 specific stations produced figures inconsistent with their underlying Form 34A photographs is a finite, examinable set. The magnitude question becomes arithmetic: do the verified station-level figures, summed correctly, change the announced outcome at the relevant level - constituency, county, or national.

Granular evidence converts a rhetorical dispute into a countable one. Aggregated evidence does not.

What does not work

A few patterns reliably fail in election petitions, regardless of how confident the people producing them are.

  • Screenshots without source. A screenshot of a result dashboard, with no link back to the underlying forms, is hearsay about a database. Courts ask for the database record, not the screenshot.
  • Aggregate claims without rows. "Our internal numbers show a difference of N" is not admissible unless the numbers and the inputs come with chain of custody. Internal confidence is a campaign concept, not a legal one.
  • Verbal reports. A field agent's testimony that a tally was wrong matters if it is corroborated by a contemporaneous artifact - a photograph, a signed form, a time-stamped log entry. On its own it is one witness against another.
  • Reconstructed timelines. A timeline assembled after the fact, using metadata the platform can rewrite, is weaker than a timeline assembled from trusted timestamps issued at the moment of each event. The difference is who controlled the clock.

Designing for what a court will accept

The implication for any platform supporting a campaign or observer organization is straightforward. The export bundle a platform produces at the end of an election cycle is the artifact that may eventually be filed. Whether that bundle succeeds depends on choices made before the first ballot is cast: station-level data model, signed image capture, timestamped chain-of-custody log, immutable export format, cryptographic signatures on the bundle itself.

These are not features that can be added in the week after a contested election. They are properties of the system, or they are not. Campaign technologists planning for the 2027 cycle will find that the most important conversation they have with any platform vendor is not about dashboards or anomaly detection. It is about what comes out of the system in the worst case, and whether that artifact is the kind of thing a Kenyan court actually treats as evidence.