Njeri chose a small café outside the university’s back gate.
“No recorder,” she said.
Wema put her phone in her bag. “Off record unless we agree otherwise.”
Njeri nodded.
For the first few minutes, she did not talk about the pilot. She asked whether Wema’s meal support had been restored.
“No.”
“I can restore the scholarship administratively. Today. If you withdraw the records request.”
Wema held her tea. “Why is the records request the price?”
“It is not a price. I am trying to stop this from escalating until everyone loses their jobs.”
“And the system remains?”
Njeri closed her eyes for a moment. “You keep imagining clean choices. Inside the office there are no clean choices.”
Then she described the timeline.
The pilot began as a fraud-control project. Donor renewal emphasized misuse rate and turnaround time. Human review created delays. Karanja reduced the manual-review threshold. Later, vendor proposals expanded the signal set to public lifestyle data and peer referrals.
“Did Karanja know it would be unfair?” Wema asked.
“I do not know. Do not invent a motive for me.”
Wema nodded.
Njeri took out one page. It was not a student database. It was a process memo with a date, version number, and policy references.
The memo said manual review had been removed for low- and medium-confidence flags to improve turnaround metrics. It also described integration into additional services as an “administrative efficiency enhancement.”
“Can I take this?”
“Not yet.”
“Is it official?”
“Internal draft. Version controlled.”
“Signed?”
“The system retains the version.”
Wema pushed the page back. “Then I do not want a stolen copy.”
Njeri looked surprised.
“There is an ethics mailbox,” Wema continued. “If you submit it there, it becomes a protected institutional record.”
“My office is small. People will know.”
“Then you decide whether to do it.”
“And your scholarship?”
“I am not trading the records request for restoration.”
“You may graduate with debt.”
“That hurts. But yes.”
Njeri folded the memo and returned it to her bag.
Wema walked away feeling almost empty-handed. She had a source’s description, not evidence she could publish.
At 23:41, Njeri sent one string.
**ETH-2026-4417**
Then: *Protected ethics inquiry accepted. Version archived.*
Wema contacted the ethics office only to confirm process. The number indicated that a record existed in a protected channel accessible to authorized reviewers. The office would not identify the submitter or discuss contents.
That was enough.
She wrote in her notes: *Inquiry is not a finding.*
Halima wanted to publish that a whistleblower had come forward. Wema refused.
“We can say an institutional ethics inquiry has opened. We do not turn the source into the story.”
Karanja later sent her a message: *Please stop encouraging staff to break confidentiality.*
Wema replied: *I have not asked anyone to provide private student data. My reporting uses public policy, my own data, consented testimony, and confirmation of an ethics process. If you identify a factual error, I will correct it.*
There was no answer.
Njeri sent one message the following day.
*Thank you for not making me the story.*
Wema replied:
*The process is the story.*
The ethics record gave her a legitimate chain, but not a shortcut. To understand how the mandate changed, she did not need another leak.
The next morning, Wema received a delayed response to her data-dictionary request. It did not provide definitions; it said the request required consultation with the vendor. She recorded the delay without describing it as obstruction. Njeri’s protected inquiry and Wema’s own request were now moving through separate channels, which reduced the temptation to make one source carry the whole case. Wema also told Halima that they would not publish the memo’s claims until an authorized review or another documentary route corroborated them. The ethics ticket proved that a disclosure existed. It did not yet prove what conclusion the institution would reach.
She needed the training materials and council records people already possessed lawfully.