Part of the AI Regulation News hub.
Kenya's draft national AI policy proposes a published Fair Pay Reference Framework benchmarking data annotation and content moderation pay to international rates
Most national AI strategies treat the annotation workforce as a footnote about jobs created. This draft treats it as a labour-standards problem and proposes a published pay benchmark calibrated to what the same work earns elsewhere.
Bottom line: A draft policy out for public participation. It binds nobody. It creates no obligation, no offence and no cause of action, and several of its measures depend on legislation that does not yet exist.
Who this affects: HR and employment counsel at business process outsourcing operators and AI data-services vendors in Kenya, procurement leads at foreign AI labs buying annotation capacity, and executives overseeing outsourced content moderation.
Issue date: The document's cover carries only JULY 2026, with no day. The ministry's public participation notice set a submission deadline of 4 August 2026, which has passed.
What changed: Nothing yet in law. The draft proposes a published Fair Pay Reference Framework for annotation, moderation and AI quality evaluation roles, plus a compliance reporting mechanism through which operators disclose pay structures against those benchmarks.
Analysis: The calibration clause is the whole point. Benchmarks set to international rates for equivalent roles attack the arbitrage that made Kenya attractive for this work, and the reporting mechanism is written to reach operators engaging Kenyan value chain workers, with international AI employers named among the parties the measure runs against.
Primary sources: Ministry public participation notice · Draft policy, full text (PDF)
- Instrument (EN)
- Kenya Artificial Intelligence and Other Emerging Technologies Policy, July 2026 (draft)
- Authority
- Ministry of Information, Communications and the Digital Economy, through the State Department for ICT and the Digital Economy
- Jurisdiction
- Kenya
- Status
- Draft, issued for public participation. 226 pages including implementation and monitoring matrices.
- Bindingness
- Not binding. A policy document, not legislation. It repeatedly describes measures the Government shall develop, gazette or promote, which is a commitment to make instruments later.
- Issue date / next deadline
- July 2026 on the cover. Comments closed 4 August 2026 per the ministry notice.
- Lead ministry for the pay measure
- Ministry of Labour and Social Protection, with the ICT ministry, the Kenya National Qualifications Authority, civil society and worker representative bodies in support.
- Primary source
- https://ict.go.ke/sites/default/files/AI%20Policy%20Doc/draft-kenya-ai-and-emerging-technologies-policy-2026.pdf
The proposal, in the draft's own words
Deep inside the implementation matrix, under a strategy headed support the development and integration of fair and transparent pay standards, the draft sets out its expected output. It is a published Fair Pay Reference Framework for AI and other Emerging Technologies value chain roles, specifying transparent pay benchmarks for data annotation, content moderation, and AI quality evaluation work, calibrated to international rates for equivalent roles.
The second half of the same output is a compliance reporting mechanism through which operators engaging Kenyan value chain workers disclose pay structures against the Reference Framework benchmarks.
Timeframe given: years two to four. Estimated cost: 30 million shillings. Lead: the Ministry of Labour and Social Protection. The monitoring matrix sets a target of at least 50 percent of these workers holding formal credentials and compliance monitored across at least 100 enterprises.
Why the draft says it is doing this
The problem statement is unusually direct for a national strategy. It lists inadequate protection of AI value-chain workers as one of the challenges the policy exists to solve, and describes annotation and moderation workers as facing unique occupational risks, including exposure to harmful content, insecure working conditions, and limited labour protections. It states that existing frameworks provide insufficient safeguards for mental wellbeing, transparency, and employer accountability.
Elsewhere the draft says Kenya's participation in global AI value chains is concentrated in lower-value segments such as data annotation and content moderation, with limited structured pathways for progression into higher-value roles. That framing, low-value segment plus no ladder, is what the pay benchmark and the credential-recognition measure are jointly aimed at.
The pay measure is one of four, and the others have teeth on paper
Policy statement 3.3.5 groups four measures for value chain workers: occupational protection instruments, recognition of experience as formal credentials, the fair pay standards, and duty-of-care standards covering harmful content exposure, mental health support, transparent contracting, proportionate workplace surveillance, and accessible grievance and redress.
The occupational protection line in the matrix goes further than guidance. Its expected output includes a compliance and enforcement mechanism applicable to all local and international AI and other Emerging Technologies value chain operators in the Kenyan market, covering minimum standards for written contracts, psychosocial support, grievance mechanisms and working conditions.
Separately, the draft's legislative agenda lists statutory duty-of-care standards for data annotation and content moderation workers, including safeguards relating to harmful content exposure, mental health support, fair contracting, remuneration, and workplace protections. That item sits in a list of things a future AI law should contain, so it is a proposal about a proposal.
What a buyer of annotation capacity should take from this
Nothing here changes a contract today. What changes is the direction of travel, and the draft is explicit that the reporting mechanism is meant to reach international operators engaging Kenyan workers, not only Kenyan-incorporated suppliers.
The practical exposure is reputational before it is legal. A published benchmark calibrated to international rates creates a number against which any disclosed pay structure can be compared, by journalists and by counterparties, well before any enforcement mechanism exists. Firms that would not want that comparison drawn have roughly the years-two-to-four window the draft itself sets out.
One more item belongs in the data strategy file, not the employment one: the draft proposes a framework to recognise and develop Kenya's domestic data annotation and labelling sector as a national data capability asset, alongside gazetted dataset documentation and provenance standards.
What we did not verify
We opened the ministry's public participation notice and downloaded and read the 226-page draft policy PDF, including the problem statement, policy statement 3.3.5, the skills and workforce narrative, the legislative agenda list, and both the implementation matrix and the monitoring and evaluation matrix rows for the pay measure.
We did not open the Kenya AI Bill, the National Digital Master Plan, the Kenya Cloud Policy, or the KNBS Economic Survey 2023 that the draft cites for the informal sector figure. We did not read all 226 pages line by line and we did not verify whether any comments submitted before 4 August 2026 have changed the text.
We do not claim the Fair Pay Reference Framework exists, that any pay disclosure obligation applies to anyone now, or that the draft will be adopted in this form. The brief we worked from dated this instrument 21 July 2026; the document itself carries only JULY 2026, and we go with the document.
Kenya is proposing to publish what annotation and moderation work should pay, benchmarked to international rates, and to ask operators to disclose their pay structures against it. None of that binds anyone today. But a published benchmark is durable in a way a policy document is not, and firms buying Kenyan annotation capacity should assume the number will exist before the enforcement does.
Source File
Download the draft policy PDF and search for Fair Pay Reference Framework. Confirm it appears in the implementation matrix with a year two to four timeframe and the Ministry of Labour as lead, that policy statement 3.3.5 lists four worker-protection measures, and that the cover page reads JULY 2026 with no day.
A published Fair Pay Reference Framework for AI and other Emerging Technologies value chain roles, specifying transparent pay benchmarks for data annotation, content moderation, and AI quality evaluation work, calibrated to international rates for equivalent roles. ยท Draft Kenya AI and Other Emerging Technologies Policy, July 2026
FAQ
Does any Kenyan employer have to disclose annotation pay today?
No. This is a draft policy. It proposes a framework and a reporting mechanism, both of which would have to be created before anything applies.
Would the pay benchmark apply to foreign firms?
The draft's expected output describes a reporting mechanism for operators engaging Kenyan value chain workers, and the parallel occupational protection measure is expressed as applying to all local and international operators in the Kenyan market. Neither exists yet.
Can I still submit comments?
The ministry notice set a deadline of 4 August 2026, which has passed. Written submissions were to go to aipolicy@moict.go.ke copied to legal@moict.go.ke, or through the online form.
What is the timeline for the pay framework?
The implementation matrix places it in years two to four of the policy period, at an estimated 30 million shillings, led by the Ministry of Labour and Social Protection.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.