AIGAH

AIGAH

AI Governance for Animals Hub
About

Bringing animals into AI governance across the EU and UK.

AI is already in the barn and in the water. Automated systems run feeding and ventilation on farms and fish farms. Cameras, microphones and sensors record every cough, every lame leg, every death. The AI layer is added on top of existing automation and records and digests all that data. But almost none of the data is reaching regulators or the public.

The AI rules now being written in the EU and the UK barely consider animals. AIGAH works to change that. We map openings for animals in those rules and provide drafts for engaging with those openings, so that advocacy organisations can file them in their own name. Our first year covers farmed animals in the EU and the UK.

Approach

AI governance for animals in three layers

We sort AI governance for animals by who is targeted by AI governance efforts. The big AI companies (“labs”) develop general-purpose AI models. Operators (farms, slaughterhouses and fish farms) run the deployed systems, which currently consist mostly of narrow AI. And last but not least, regulators can be targeted, with the aim of getting them to use AI to enforce existing animal welfare law.

GPAI models

What general-purpose AI (GPAI) models, such as those behind chatbots, must take into account. This is part of what AI researchers call alignment.

Who is targeted

Big AI companies that develop general-purpose AI models.

Current situation

In the EU, the Code of Practice for general-purpose AI (the GPAI Code) names “non-human welfare” as an example of a systemic risk. It does not define what this means, and it does not require companies to test for it or report on it.

What we ask for

Criteria for what counts as harm to animals, and tests that AI companies run on their models and report on.

Openings

A first review of the GPAI Code is expected in 2027, and most of the leading AI companies have signed the Code.

Deployment

What operators must record about animals, and what they must disclose.

Who is targeted

Farms, slaughterhouses and fish farms that use cameras and sensors, and automated systems for feeding and climate control.

Current situation

No rule says what these systems must record about the animals, or who gets to see that information. None of the areas that the EU AI Act lists as high-risk concerns animals.

What we ask for

Disclosure of which systems are in use and on what share of farms, how they are set up and what they have found, measured against welfare indicators that are set independently.

Openings

In the EU, new farm animal welfare laws are being prepared, and they are to include “practical on-farm welfare indicators”. A Commission proposal on laying hens and broilers is expected by the end of 2026, and one on pigs by mid-2027. In the UK, there are openings through the Animal Sentience Committee and the farm assurance schemes.

Enforcement

Using AI to enforce animal welfare law that already exists.

Who is targeted

Regulators and enforcement bodies, which could use AI to check whether animal welfare law is being followed.

Current situation

CCTV is mandatory in slaughterhouses across the UK and in Spain. In England the official vet must view at least 15 minutes of footage a day, which is a small fraction of what the cameras record. AI that screens such footage exists, but where it is used now, its alerts go to the company and not to the regulator.

What we ask for

AI analysis of the footage that is already being recorded should go to the regulator, so that existing animal welfare law can be enforced.

Openings

Because the law and the footage already exist, we think this is where a first visible result is most likely.

One common ask that we have for all three layers: information about animals that is relevant in these systems needs to reach regulators and the public.

Across all three layers: the hub

  • Research. We map where EU and UK rules leave animals out (the regulatory gap map) and collect cases in which automated systems have harmed animals (the evidence bank).
  • Policy engagement. For each opening we write briefs, templates and consultation responses that advocacy organisations can use.
  • Field-building. We bring together organisations and people who can take this work forward, through workshops, mentoring and our newsletter.

Read our theory of change →

Why now

Key dates

It is far easier to shape a rule while it is being drafted than to change it once it is in force. Several of the rules that matter for animals are being drafted now.

AIGAH activities Rules being drafted or reviewed

  1. July 2026

    Our response to the EU AI Office's consultation on classifying high-risk AI, and a guide that helps others respond within the consultation's scope. Read the guide →

  2. September 2026

    Talk at EAGxOxford: “AI governance for animals – why and how”. Slides (PDF) →

  3. October 2026

    A submission to the UK Animal Sentience Committee on AI and the government's planned regulatory sandboxes.

  4. 30 October 2026

    Workshop in London with researchers and advocacy organisations.

  5. Autumn 2026

    Our regulatory gap map is published on this site.

  6. End of 2026

    Proposal from the European Commission on laying hens and broilers, with on-farm welfare indicators.

  7. By mid-2027

    Proposal from the European Commission on pigs.

  8. 2027

    First review of the EU's Code of Practice for general-purpose AI (expected).

  9. August 2028

    Report from the European Commission on whether new areas should be added to the AI Act's list of high-risk areas.

Current projects

Regulatory gap map: EU and UK

Publishing in autumn 2026

This map shows where EU and UK rules on AI leave animals out. It covers around 30 such intervention points, and for each one it names the legal instrument that could close the gap and a way of acting on it. We are building it as a shared resource for advocacy organisations.

Operational criteria for “non-human welfare”

In progress, with a draft for expert review in autumn 2026

Non-human welfare is named in the GPAI Code of Practice as an example of systemic risk, but the Code offers no definition and no method for assessing it. We are developing criteria with which this risk can be defined and assessed, and we plan to circulate them before the Code's first review, which is expected in 2027.

UK: a submission to the Animal Sentience Committee

Submitted October 2026

Laura Thomas-Walters, working with AIGAH, has made a submission to the Animal Sentience Committee. It asks the Committee to examine the regulatory sandboxes that the government plans for AI and other new technologies, meaning pilots in which existing rules can be relaxed for a limited time, and what these sandboxes mean for animals. In particular, it asks that animal welfare be named among the protections that a sandbox cannot modify or disapply. UK organisations can co-sign the submission, or they can file their own version.

Co-sign or file your own →

EU consultation: high-risk AI classification

Submitted July 2026

In 2026 the EU AI Office consulted on draft guidelines for classifying high-risk AI systems under the AI Act, and the final guidelines are expected at the end of the year. Failures of AI in animal contexts, for example in the automated climate control of livestock housing, fall within the risks that the guidelines address, but the draft contains no livestock example. By the deadline on 23 July, AIGAH submitted its own response and published a guide to responding within the consultation's scope. Four organisations and two researchers used the guide for their own responses.

Our response guide →
People
Dr Henrike Gätjens

Dr Henrike Gätjens

Founder and director

I founded AIGAH in 2026 to bring animals into the conversations and decisions happening now in AI governance. Before that, I did a PhD in the history of science, facilitated the Sentient Futures AI×Animals Fellowship, and helped Max Taylor with research for his forthcoming book AI, Animals, and Us (HarperOne, 2027). I'm also a Research Associate at LSE, working with the Jeremy Coller Centre for Animal Sentience. I'm based in Cambridge, UK.

Benjamin Hong

Benjamin Hong

Researcher

I'm a software engineer moving into AI governance. At HUX AI I worked on evaluating guardrails and risk controls for LLMs and AI agents, and the Anima International Fellowship brought me to animals. At AIGAH I'm working on how benchmarks for AI risks to animals can be used in regulation, starting with the EU's Code of Practice for general-purpose AI.

Advisors

Advisors give their input in a personal capacity; AIGAH's positions are its own.

Working with AIGAH this autumn

AIGAH mentors three people from the Sentient Futures Project Incubator from September to November 2026. Their projects deal with the UK, with incidents in which automated systems harmed animals, and with the Philippines. In all three projects, the first ask is transparency.

Get involved

If you work at an animal advocacy organisation

Once the gap map is online, it will show the openings in EU and UK rules, with briefs and templates that you can adapt and file in your own name. If you are in the UK, your organisation can co-sign the submission to the Animal Sentience Committee. We would also like to hear which consultations you follow.

If you work on AI governance

You can include non-human welfare in work you are already doing, for example in a risk taxonomy, an evaluation, an impact assessment or a consultation response. Animals are affected by AI but cannot speak for themselves, so they are a useful test of whether a framework covers everyone it affects.

If you fund or research this field

Our theory of change describes what we expect to change by 2030 and how our work should contribute to that. If you would like to talk it through, please write to us at hello@aigah.org.

Stay in touch

We send occasional updates on our research, policy work and events.

Or write to us: hello@aigah.org