Project summary

We are building a scalable AI risk literacy platform designed to reduce harmful misuse, overreliance, and manipulation risks associated with increasingly capable AI systems.

Our focus is not general education, but targeted behavioral intervention—helping users recognize dangerous capabilities, avoid unsafe use patterns, and understand real-world risks.

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

• Goals

  • Train 100,000+ users to recognize and avoid high-risk AI misuse patterns

  • Reduce overreliance on AI in high-stakes contexts (health, legal, financial)

  • Build a dataset of user misunderstandings and failure modes

  • Contribute to AI safety evaluations and alignment research

• Dataset for AI safety evals

  We will collect structured datasets including:

  • Common failure modes (hallucination reliance, unsafe delegation)

  • Prompt patterns leading to harmful outputs

  • User misinterpretation of model capabilities

  • Behavioral data on overreliance in simulated high-stakes scenarios

  This dataset will be formatted for direct use in AI safety evaluations and red-teaming pipelines.”

• “As frontier AI systems rapidly scale in capability and accessibility in 2026, misuse risks are increasing faster than public understanding. There is a narrow window to intervene before unsafe usage patterns become widespread.”

• “We will make datasets and findings available to AI safety researchers, evaluation teams, and alignment organizations. We aim to collaborate with groups working on model evaluations, red-teaming, and safety benchmarks.”

  we measurably change behavior that reduces risk”

• “This project addresses a critical gap: while frontier AI capabilities are rapidly advancing, there is currently no large-scale infrastructure for measuring and reducing harmful user interaction patterns with these systems.”

How will this funding be used?

Breakdown (~$250K)

Risk Intervention Infrastructure ($120K)

• Development of core platform for large-scale AI misuse prevention

• Interactive simulation systems for high-risk AI scenarios

• Adaptive AI feedback system to correct unsafe user behavior in real time

AI Safety Curriculum & Scenario Design ($50K)

• Development of structured modules on:

  • AI misuse patterns

  • Overreliance in high-stakes contexts

  • Manipulation and adversarial risks

• Creation of interactive simulations replicating real-world failure cases

• Continuous iteration based on observed user behavior

Targeted Distribution & Data Collection ($40K)

• Distribution to high-risk and high-impact user segments

• Partnerships with organizations, educators, and communities

• Systems to capture structured data on:

  • User misunderstandings

  • Failure modes

  • Behavioral responses to AI systems

Infrastructure & Operations ($25K)

• Hosting and infrastructure to support scalable data collection

• Secure storage and processing of behavioral datasets

• Tooling for analysis and integration with AI safety research workflows

Contingency & Rapid Response ($15K)

• Flexibility to respond to emerging AI risks or opportunities

• Team & Track Record

  Our team combines experience in building scalable educational systems, technical development, and AI-focused content with an emphasis on real-world usability.

  Execution Track Record:

  • Built and deployed 100+ structured AI guides currently used by thousands of monthly users

  • Established organic distribution channels that allow us to reach real users interacting with AI systems at scale

  • Developed a high-performance, accessible platform optimized for clarity and rapid iteration

  Unique Advantage:

  • Direct access to real user interaction data at scale, enabling us to capture how people actually misunderstand and misuse AI systems in practice

  • Ability to rapidly convert observed failure patterns into structured learning interventions and datasets

  As a nonprofit, we are focused on open, transparent, and broadly accessible infrastructure aligned with reducing real-world risks from advanced AI systems.

Expected Outputs

• Dataset of 100,000+ user interactions with AI systems highlighting misuse patterns

• Taxonomy of common AI misunderstanding and failure modes

• Open-source simulation modules for AI safety education

• Public reports summarizing behavioral risk trends in AI usage

Risks

• Failure to capture meaningful or high-signal misuse data

• Insufficient behavioral change in user interaction with AI systems

• Rapid evolution of AI capabilities outpacing curriculum relevance

• Limited integration with existing AI safety research and evaluation pipelines

Mitigation

• Focus on high-risk, high-signal scenarios (e.g. overreliance, unsafe delegation, manipulation)

• Continuous iteration using real user interaction data

• Rapid deployment of updated modules in response to frontier model changes

• Structuring all collected data for compatibility with AI safety evaluation and research use

If it fails

Even in failure, the project will produce:

• A structured dataset of real-world AI misuse and misunderstanding patterns

• Insights into how users interact with increasingly capable AI systems

• Open-source tools and infrastructure that can be reused by AI safety researchers and educators

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

Approximately $10,000 in self-funded development and operations over the past 12 months.

This funding has supported initial platform development, content creation, and early user acquisition—demonstrating our ability to execute efficiently and build meaningful infrastructure with limited resources.