Work

I help organisations make AI rules they can enforce and check.

Plenty of AI rules exist on paper. Few can be enforced, and fewer can be checked by anyone outside the room. I work at the point where a principle has to become a clause, a control or a screen. I've spent 15+ years turning regulation into systems people can run.

Where I'm brought in

Before a launch touches people's data or likeness

I check that consent and authorisation hold up in law and in the interface. I've led consent work for a national digital programme, and I have documented what happens when image tools ship without safeguards.

When a board must answer for AI

I help boards see what they are accountable for under the EU AI Act and what to ask management. I sit on boards myself, and I chair an audit, risk and compliance committee.

When rules or policy are being drafted

I help governments and funders write policy that can be put into practice and checked. I've led AI readiness assessments for national programmes in several countries, and trained 45 regulators.

When terms promise more than the product does

I read the contract, the policy and what the system actually does, side by side. I also read what users report about how it performs, and weigh that against the standards of each industry it could be used in. It is careful, detailed work. Then I turn it into something a team can use, like the libraries of risks, products, models and policies that I lead.

When many actors need to move together

Some work needs people with different interests and goals to pull in the same direction. I help them agree on shared values and a common direction. Then I design how they communicate, deliver and measure impact together.

When a non-profit wants to use AI responsibly

This is some of my favourite work. A non-profit comes to me wondering where AI fits, and we start with them: the communities they serve, the data they hold, the team they have and what they can actually sustain. Then we find the places where AI genuinely helps. I build a small first version they can try for real. Once it works, I help them grow it responsibly with open-source tools they can own.

How I think

Platforms should have to prove their systems prevent harm.
Removing harm once someone finds it comes too late for the people already hurt. I watched requests for nonconsensual images of women pile up for months before the backlash began.
Prevention belongs in the law and in the code.
That means input filtering, independent audits and licensing conditions that make prevention a legal and technical requirement, before a product reaches people.
Consent counts when someone can check it.
I keep coming back to consent. Who agreed, to what, and who can check. If nobody can check it, nobody can enforce it.

Stories

2025 to 2026

Counting what an AI image tool was asked to do

EvidencePublic writing

Women were posting photos on X and, within minutes, other users were tagging Grok to strip them. It was happening at scale, and nobody was counting it.

I decided the most useful thing I could do was count, carefully, and publish the data so anyone could check it. Between June 2025 and January 2026 I documented 565+ requests to Grok for nonconsensual intimate images.

I released the dataset openly and wrote a Guardian column on what the law was missing. Regulators in six jurisdictions responded. The dataset stays public, and I still use it to argue that platforms should have to prove their systems prevent harm.

See the dataset →

2023 to 2024

Consent for a national digital programme

Consent

A country was building ways for public bodies and private companies to share citizens' data, inside a €5.25M facility. If people's consent could not be checked, the whole programme would rest on trust alone.

I led the consent work. I chose to test the risks before real data moved, so I used synthetic data to see where consent could fail.

I co-designed workshops, policy toolkits and readiness scorecards for the people who would run the system. The work ended in recommendations on how public and private data sharing should be governed, so that consent could be shown and checked.

2022 to now

A risk library every client relies on

AI rules into toolsTeams and knowledge

Teams governing AI need to know what can go wrong with the models and products they use. Most start from a blank page, and each one guesses differently.

I lead the library at the centre of an AI governance platform. I set the review standards and the release cadence, and I manage a team of five.

The library now covers 400+ risks and 200+ models and products. Every client of the platform uses it to check their own systems. It keeps growing with each release, and it is built so the team can maintain it without me in the room.

2020 to 2022

A procurement Delivery Unit built from scratch

Teams and knowledgeTraining and convening

A state government wanted public procurement that people could follow and check, on a $2.5M project. There was no team to do it.

I built the Delivery Unit from scratch. I judged that the skills had to sit with the state's own staff, so I trained everyone. I recruited 10 people and trained all 25 staff, including on freedom of information and procurement.

The unit worked across 17 local government areas. The training materials I wrote gave staff something to return to after I left.

2018 to 2021

A policy evidence portal

Teams and knowledgeEvidence

A national policy programme was producing a great deal of work, and it needed a way to keep and share what it learned.

I built the knowledge team and set reporting standards, so every output could be found and trusted. I ran 26 case and tracer studies and reviewed 112 reports.

The result was a portal of 1,000+ policy outputs that anyone on the programme could search. It turned years of scattered reports into a record people could use.

Recognition

Nanaaaa!!! I was searching for an expert and someone was describing this amazing AI governance person who broke down what they needed to do without buzzwords and who understood civil society. So I asked to be introduced and they sent me your LinkedIn. I screamed that I know you!!!! I'm so ashamed I didn't look inwards. Kai! Please we need you

Lead, UN project

She coordinated working groups, teams, consultants, stakeholders, etc. ... She is often willing to go the extra mile in providing support to colleagues and maintained an open approach policy. She’s able to negotiate reasonable trade-offs with stakeholders in order to get the job done.

Former Director-General, public procurement agency

Allow me to thank you specially for the amazing coordination work you have done behind the scenes and on the scenes. You not only hold our hands but honestly guiding us onto the right path as you cheer us up in ensuring that community lead the change that most matters to them.

Founder and Director, community-led development NGO

You are an exceptional writer! WOW! The article was beautifully expressed and is a very accurate summary of the way the learning affected the community at our school.

School project leader

[The applicants] did very well in their [pitches] regarding any ethical issues and aspects! Thanks again for the great work!

AI Policy Manager, AI governance company

I always enjoy working with you because you trust and believe in me!

Programme advisor and colleague

Ways to work together

Retainer advisory

Steady advice for a team building or governing AI, there when the hard calls come up.

Board session

A focused session with your board on risk management, strategic positioning and governance frameworks for AI.

Independent review

A close read of a product, policy or set of terms, with clear findings on where they fall short.

Keynote or briefing

A talk or briefing on consent, platform accountability or what the evidence shows.

I don't take on work that asks me to sign off on something I haven't been able to check.

Start a conversation

For procurement

Nana Nwachukwu smiling, wearing clear glasses
Nana Nwachukwu

Short bio (50 words)

Nana Nwachukwu is a technology lawyer and AI governance expert. For 15+ years she has turned regulation into systems organisations can run. She works on making consent, authorisation and transparency in AI enforceable. She researches accountability in AI systems at Trinity College Dublin and was an Affiliate Fellow at Harvard's Berkman Klein Center.

Longer bio (150 words)

Nana Nwachukwu is a technology lawyer and AI governance expert. For 15+ years she has turned regulation into systems organisations can actually run. Her focus is making consent, authorisation and transparency in AI enforceable.

She leads a library of 400+ AI risks and 200+ models and products used by every client of an AI governance platform. She has led AI readiness assessments for national programmes in several countries, and led digital consent work for a national digital programme.

She researches accountability in AI systems at Trinity College Dublin's AI Accountability Lab. Her dataset of 565+ requests to Grok for nonconsensual intimate images drew responses from regulators in six jurisdictions, and she wrote about it in the Guardian. She was an Affiliate Fellow at Harvard's Berkman Klein Center and sits on the advisory board of the Digital Democracy Initiative. She is Nigerian and lives in Europe.

Download CV