Project summary:

~1B people use AI daily. AI agents are coding our healthcare systems.  No accessible AI nutrition label to make an informed choice.

Fix:  1. Standardised AI nutrition labels, readable by everyday consumers & AI agents

2. Global, open registry of these labels. Developed in collaboration with ISO, the IASEAI, and the United Nations

I've been invited to present my recommendations on "AI Nutrition Labels For Everyday Consumers & AI Agents" at the UN Global Dialogue on AI Governance (July 6th-7th), and the AI For Good Summit (July 7th-10th) in Geneva, Switzerland.

The journey of the AI Nutrition Labels framework looks like this (better depicted by a picture): 

1. Apr, 2025: “AI Nutrition Labels for everyday consumers”, developed in collaboration with ISO expert (Harm Ellens), ex-US EPA (Nathaniel Burola): Started as a BlueDot Impact AI Governance project, published on Medium

2. Jan, 2026: Got invited to IASEAI 2026 based on it

3. Feb, 2026: Presented this to multiple experts at IASEAI (incl. researchers at Mistral working on benchmarks, Prof. Stuart Russell, Amir Banifatemi). I received two major pieces of feedback from them : 

• Make this use-case or sector specific, like different labels for healthcare, child safety, etc. 

• Extend this framework for AI Agents

Note: Super grateful to have received IASEAI travel funding from Manifund (major funding by @JueYan ❤️)

4. May, 2026: Presented my IASEAI-feedback-informed recommendations at one of the online stakeholder consultations for the UN Global AI Dialogue for AI Governance. After my presentation, I received LinkedIn DMs from stakeholders in the meeting, including members of the UN's Independent International Scientific Panel on AI, supporting my recommendations.

5. Jun, 2026: Got invited to the First UN Global Dialogue for AI Governance

6. Today: Looking for a Travel + POC grant to take it further

Why AI Nutrition Labels For Everyday Consumers?

When you buy a loaf of bread, you don't need to be a food scientist; you read the label for calories, allergens, and within seconds, you make an informed choice. Yet, for AI, we expect a teacher in Nairobi to read a 60-page technical "model card" to know if a tool is safe for her students. This isn't a literacy problem; it's a governance problem.

Why extend this to AI Agents?

AI coding agents are independently selecting the models, APIs, and tools that will power tomorrow's healthcare systems and educational platforms, with no way to ensure they are choosing components that are safe for the intended domain, unbiased, or environmentally responsible. For example, an AI agent building a maternal health app for rural India selects a language model, an imaging API, and a translation tool. Each choice shapes who that system will serve, and who it will fail. Right now, those choices are not well-informed. The stakes are now exponential because AI agents are building our world.

What are this project's goals? How will you achieve them?

My attendance at the First UN Global Dialogue on AI Governance and the following “AI For Good Summit” has two objectives:

1. AI Governance:

- Present my recommendations backed by a working POC to relevant stakeholders like Seizo Onoe, Sasha Luccioni, Sergio Mujica, Philippe Metzger, and others at the Dialogue and the following “AI For Good Summit”

- Join or form a working group on AI transparency standards and registries, and push this toward an adoptable global standard and usable registry integrated into the existing AI Standards Database. This work continues post the Geneva events

2. AI risk literacy:

- Record in-person podcast episodes with 

a. Yoshua Bengio (already in progress with his Comms team: Claire/Marc)

b. Sasha Luccioni (WIP)

c. Seizo Onoe (WIP)

d. Amandeep Gill (WIP)

and bring their voices to my audience in India, where 55% of my listeners are based, and where credible AI safety and governance voices are still nascent

- The relevant stakeholders at the Dialogue and the summit can be potential funders for our AI safety workshops for Indian middle schoolers (6th–8th Grade), a talent pipeline we've been building and are actively seeking grants for

Before going to Geneva, I need something concrete to put on the table. Showing up to a UN Dialogue on AI Governance with a framework but no working proof-of-concept is the difference between "interesting idea" and "here's what this looks like in practice." So in approx. three weeks before July 6th, I'm building:

1. A technical POC that makes the framework AI agent-readable and actionable with the following:

    - JSON schema (v0.1-draft) for the AI Nutrition Label

    - A mock registry with 5–6 dummy model labels

    - REST APIs to query the registry

    - A domain-specific AI agent demo (healthcare, education, finance): given a task and constraints, the agent queries the registry and makes an informed model choice with reasoning

2. And a detailed, actionable governance recommendation document based on my oral input at the UN Stakeholder Consultation. It addresses:

 - A global standard for accessible AI disclosures regarding safety, bias, and environmental impact

 - A global, open registry of these labels. This allows a mother to know if a tool is safe for her child. Labels that tell an AI agent procuring components for a healthcare system: Does this model meet the safety threshold for that use case? Is it trained on data representative of the population it will serve? 

How will this funding be used?

My grant request: 

1. Travel grant to present my recommendations on "AI Nutrition Labels For Everyday Consumers & AI Agents" at the UN AI Dialogue (6th, 7th July) and the following AI For Good summit (7th-10th July)

2. Grant to create a technical POC extending the AI Nutrition Label framework to the AI Agents + Governance document to present there. The POC is what turns a conversation into a working group

Funding Goal Amount = $8,466.86

1. Travel (flights, accommodation, meals, local transport): ~$5,797.94

2. Visa and travel insurance: ~$195.29

3. Technical POC + hosting/API credits + Podcast toolkit: ~$1,369.27

4. Contingency (15%): ~$1,104.37

Min. Funding Amount = $5,764.89

1. Travel (Accommodation, meals): ~$3,643.68

2. Technical POC + hosting/API credits + Podcast toolkit: ~$1,369.27

3. Contingency (15%): ~$751.94

More details here

Note: The budget is higher than my previous travel grant for two reasons:

1. It includes attending two major events in Geneva, Switzerland

2. Due to the summer time and the "AI For Good" summit, the accommodation is very expensive and hard to find. You can find more details here.

Also, for the minimum funding amount, I've applied to the Emergent Ventures Grant to support the remaining amount ($2,701.97). Applied but haven't received the decision yet. If the Emergent Ventures Grant doesn't come through, this project would be totally dependent on Manifund.

Who is on your team? What's your track record on similar projects?

Aashka Patel: Self-funded through personal savings and Manifund support, I've taken the AI Nutrition Labels framework through this journey from a BlueDot course project to a UN Dialogue invitation in fourteen months.

Apart from this, self-funded through personal savings, I've built (earliest to latest):

• AI Safety comics used in Bay Area schools for K-12 education in 2023

• My Bug Bounty work at Anthropic, identifying CBRN threats, keeps me grounded in frontier AI risks

• Podcast episodes featuring White House AI policy experts, US Senate advisors, sustainability researchers, AI Ethics Professors, etc, building an informed community excited about AI's benefits while aware of its risks

• "AI Ki Adalat," a video series examining real AI incidents through courtroom storytelling to inspire Indian youth toward safety careers, not doomerism

• AI Governance corporate and college sessions at HSBC India, DataCamp, etc

• "Mummy Padhegi AI", an AI risk literacy workshop series for Indian mothers, because as Napoleon said, "Give me an educated mother, and I shall promise you the birth of a civilized nation"

• AI Safety workshops for Indian middle-schoolers (6th-8th Graders)

Amir Banifatemi may serve as an informal advisor; still waiting to hear back from him (fingers crossed).

Otherwise, it will be entirely me for the POC work, and then the relevant above-mentioned stakeholders will get involved at the Dialogue and the Summit.

What are the most likely causes and outcomes if this project fails?

Most likely cause: Insufficient funding, forcing me to either skip Geneva entirely or show up without the POC.

Outcomes if unable to attend:

• I lose the opportunity to present this framework at the first UN AI Dialogue and to high-profile stakeholders (Seizo Onoe, Sasha Luccioni, Sergio Mujica, Philippe Metzger) that would actually matter because the working groups forming at this first UN AI Dialogue will shape AI transparency standards for the next decade

• I lose the chance to record in-person podcast episodes with Yoshua Bengio, Sasha Luccioni, and others, which would bring credible AI safety and governance voices to my audience in India, where this discourse barely exists

• I lose direct access to potential funders for our AI safety workshops for Indian middle schoolers, which we're actively seeking grants for

How much money have you raised in the last 12 months, and from where?

Apart from the Manifund funding for my IASEAI travel, I haven't received any external funding for this project.

Other grant applications:

• Emergent Ventures Grant (Applied but haven't received any decision)

• BlueDot Impact Rapid Grant (Applied but rejected)

I believe in the idea of "Vasudhaiva Kutumbakam - The whole world is like one (my) family" and don’t want to jeopardize the very existence of my family due to unsafe AI advancement 🌻

Manifund has been greatly supportive of this project, and any further support would mean the world for this project ❤️

Thanks a ton in advance,

Aashka Patel 🤗