AI Safety Literacy for Frontline Institutions

Embedding AI safety literacy into AI Adoption workshops and creating AI policy for an civil society or low-resource organizations.

Project Details

Everyone is deploying AI right now, including high-stakes, low-resource institutions where AI is supposed to be "the great equalizer."
There are plenty of AI adoption courses, but they are built on organizational efficiency, and basic prompting. Some are on governance, ethics, and policy, but rarely are they on technical safety.
More and more these orgs (or rather people in the orgs) are moving towards shadow AI, not understanding data privacy or training, or maybe don't have the tools for discernment with hallucinations and confidence. These orgs are also moving towards agentic tools, and they don't understand failure modes, loss of control, etc.
I already work pro-bono with several mission-driven organizations with low-resources to build their AI capacity to increase impact. These organizations are justice organizations, working with vulnerable populations (immigrants, the incarcerated, and formerly incarcerated), and I want to stress the importance of safety with these populations and their data.
This project will fund building a safety-literacy module, and work hands-on, pro-bono for these justice organizations and other non-profits in need. It will also fund community-building initiatives, such as events and mentorship, to bring together these organizations with each other and other AI safety engineers and policy makers to deepen their community and learning.

Civil society's AI involvement is often in forms of getting them to use AI, for grants, for efficiency. There are many field building pipelines like BlueDot, but those are for people getting into technical AI safety, not necessarily those that need guardrails in their AI pipelines and need to learn about.
Safety (and sustainability) is a luxury for many of these Civil Society organizations to think about, but it shouldn't be. It should be a basic standard.
Right now, the biggest challenges aren't that no one is using it, it's they are all using it on free plans or non-enterprise plans, not understanding the risks, and accepting the outcomes as they are. The staff are adopting the tools faster than anyone is tracking the issues and failure.
Organizations adopt AI policies or learn about AI once (or only have the capacity to attend training once a year), and this field and safety news changes daily, and needs someone to help them understand this is a process, and really help them update their policy, safety measurements, and implementations constantly.
This project will give a repeatable curriculum module, living framework toolkit that helps them understand what to update and when. More people will be trained and educated.

Team Member

Applied

Minimum

$20,000

Ideal

$45,000

How the money will be spent

MIN:

Time working probono designing the safety-literacy modules for these organizations, delivering hands-on workshops and assists to these organizations
Stipend for a technical safety person to co-teach a session/collaborate with these orgs
Logistics costs: scheduling, materials, space costs, etc.

IDEAL:

Building out a system that automatically looks at the latest safety, policy, and technical issues and helps ping, educate, and assist these organizations with changes in this fast moving AI world
Expanding the pilot to more frontline organizations, testing whether these modules holds up with partners I don't previously already have relationships/deep trust with
Stipends for more technical and policy oriented co-curriculum designers and facilitators

No comments yet. Be the first to share your thoughts.