Reducing risks of AI misuse

AI Misuse Lab investigates how humans might use advanced AI to cause catastrophic harm.

Capability

Voice cloning

Capability

Image generation

Capability

Video generation

Programme+

Hackathons

Our focus

Forecasting, modelling and defending against AI-enabled fraud and catastrophic harm.

LIVE

Real-time impersonation

Capability

Synthetic identity

Capability

Agentic scams

Programme+

Policy Fellowship

Sectors
FinanceHealthEducation

Where harms could be most serious

• About us

A non-profit studying how people could turn advanced AI into tools for fraud and catastrophic harm

• What we study
6

AI capabilities, from voice cloning to agentic scams, studied across three key sectors.

Misuse vs alignment

Misuse is a problem of human intent.

AlignmentWhether an AI system pursues the goals its designers and users intend.
MisusePeople deliberately using AI to cause harm. This is our focus.
Sectors
3
FinanceHealthEducation
Cross-cutting topics
Detection evasionOpen-weight safeguards
2

Forecasting, modelling and defending.

We focus on the ways advanced AI capabilities, such as voice cloning and image and video generation, could be used to perpetrate fraud and other catastrophic harms across finance, health and education.

Step 1

Forecast

Identify realistic scenarios in which people could use advanced AI to cause harm.

Step 2

Model

Work out how those scenarios would unfold, who they would affect and how badly.

Step 3

Defend

Build defences, evaluations, tools and policy ideas to counter them.

Voice cloning

Cross-cutting

Detection evasion

How misuse is hidden from the systems built to catch it. Applies to every capability we study.

Cross-cutting

Open-weight safeguards

How protections hold up when model weights are openly available. Applies to every capability we study.

Researchers and mentors, working on AI misuse.

Our programmes bring talented researchers together with experienced mentors to work on AI misuse.

Programme

Hackathons

Test AI agents in controlled environments and find where they fail in healthcare, education and finance. Every finding is reviewed before it is published.

TestReproduceReport
See the brief
Programme

Policy Fellowship

Turn technical findings into practical recommendations for governments, labs and companies.

GovernmentsLabsCompanies
Express interest
• Team

The people behind the lab.

A small team working with researchers and mentors on how advanced AI could be misused, and how to stop it.

Nathaniel Opoku

Nathaniel Opoku

Founder

Works on AI governance consulting and AI safety fieldbuilding across Africa and emerging markets. He also founded CoLoG Africa and is based in London.

LinkedIn
Stanley Nnaka

Stanley Nnaka

Co-founder

An AI and blockchain researcher working on how new technology gets built and used responsibly.

LinkedIn

Work with us directly.

Join the team or mentor a fellow. To take part in a hackathon or the Policy Fellowship, see Programmes above.

Frequently asked questions.

Help close the gap.

Between how AI is meant to be used and how it is actually used. Join a hackathon, mentor a fellow or work with us.

Get updates from the lab.

New hackathons, fellowship rounds and research, sent to your inbox.