Project
Database
The AI Safety database is a work in progress, built from public sources. Missing, or out of date? You can list your project yourself. Feel free to send us ideas or feedback!
| Name | Type | Tags | Endorsed | Team | Raising $ |
|---|
Newsletter
| Name | Type | Tags | Endorsed | Team | Raising $ |
|---|
Showing 51-100 of 113·Sorted by endorsements
| Name | Type | Tags | Endorsed | Team | Raising $ |
|---|
| Project | Research | +1 | - | ||
| Project |
Individual |
+1 |
| - |
| Project | Field-Building | - | - |
| Project | Conference | - | - |
| Project | Research Lab | - | - |
| Project | Individual | - | - |
| Project | Network | - | - |
| Project | Individual | - | - |
| Project | Media | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Training | - | - |
| Project | Media | - | - |
| Project | Newsletter | - | - |
| Project | Tooling | - | - |
| Project | Tooling | - | - |
| Project | Standards | - | - |
| Project | Research | - | - |
| Project | Platform | - | - |
| Project | Platform | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Research | - | - |
| Project | Research Lab | - | - |
| Project | Individual | - | - |
| Project | Platform | - | - |
| Project | Tooling | - | - |
| Project | Media | - | - |
| Project | Field-Building | - | - |
| Project | Think Tank | - | - |
| Project | Individual | - | - |
| Project | Education | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Individual | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
| Project | Research | - | - |
Testing whether the safety circuits in language models actually work in Igbo (Nigerian Language) and building evaluations that distinguish real refusals from accidental ones.
consolidate, extend, and manually validate an AI-alignment field map, close gaps, perform backtests, post Metaculus questions, and write LessWrong posts about the map.
Requesting $750 to fund travel, lodging, and expenses for an AI Safety Nepal representative to attend Swiss AI Safety Days 2026 in Zurich to network and learn field-building practices.
Ensuring our ability to grow in a sustainable manner
An interdisciplinary research group based in Oxford for character training, model pyschology and multi-agent interaction dynamics
Flights and visa sponsorship for a first-time presenter from India, sharing research on whether LLMs behave differently when they think they are being tested.
PauseAI Canada organizes media outreach, public events and grassroots advocacy to secure Canadian support for an international treaty pausing frontier AI training.
Funds Parveen Sheikh to travel from India to present an AI safety governance poster at Swiss AI Safety Days and attend EAGxBerlin for networking and collaborations.
Produce a 4-episode technical AI alignment podcast interviewing researchers about their agendas, funding equipment, travel/space, and the host’s time.
Help me travel to Sydney to present my accepted NeurIPS 2026 research paper. I need help with visa and flight costs.
Research and write a history-of-science analysis of persistent disagreement (e.g., caloric vs kinetic heat) to draw lessons for current AI alignment disputes and progress.
NeurIPS 2026 Workshop on Evaluation of Interactive Agents (IAEval), AI for Verifiable Coding (VERICODEGEN)
Help gain a technical understanding of AI Engineering to enable effective check and challenge AI safety using a technical lens
Millions of people watch my AI safety videos each month, but so far my channel has cost me more money than I’ve made with it.
Expand "Humanity Tomorrow" website with a multilingual AI Risks & Mitigations Map connecting risks, causal pathways, mitigation strategies, remaining gaps, and project ideas, while maintaining the existing AI Risks FAQ.
Open-source hidden state control telemetry, reading and steering the black box, to prevent latent AI collapse before output, protecting Man and Beast
In-memory latent trajectory auditing for 0-token prefill eviction and 1–3 token circuit breakers across 70B LLMs.
Finish an open AI-disclosure protocol (GTC v1.0) and publish an open method for checking whether an agent's action can be undone
Machine-verified core (190 Lean 4 modules, 0 sorryAx) and GRPO-aligned model achieving +15.6% accuracy with a 37.2% reduction in reasoning tokens.
An open observatory of the data behind the AI buildout: which public records show where data centers and the power systems around them are, how open and checkable those records are, and what is still missing.
red teaming and educational ai agent platform
IASEAI'27 standards applicant. DS24 is a deterministic firewall for AI agents. I built it to cover 4 attack vectors: input, output, collusion, and sandbox escap
We will test whether AI assistance that performs developing cognitive operations weakens adolescents’ later independent judgment, and develop evidence-based safeguards that keep human reasoning meaningfully in the loop.
A two-phase study of context-induced persona-pattern uptake: first establishing reproducible behavioral contrasts and boundary conditions, then testing what mechanistic signatures distinguish those states and whether candidate mechanisms can be causally manipulated.
Establishing Nepal's biotechnology research infrastructure to develop genomic solutions for climate-resilient agriculture, strengthen food security, and advance locally led scientific innovation.
An open-source, model-agnostic governance layer that uses constitutional state, trajectory memory, and mathematical control mechanisms to govern LLM and AI-agent behavior at runtime.
Consumer Awareness Platform for AI Risks
ClauseHound builds a self-hosted privacy-first legal AI node using open-source LLM agents, secure networking, retrieval/search tools, and RL/eval loops to reduce hallucinations and fit law firm workflows.
An independent Australian podcast raising public and policymaker understanding of catastrophic risks from transformative AI — hosted by an ethicist and an AI-governance practitioner in an under-served region for AI safety.
Funds BASE Research’s executive team travel to EA Global NYC 2026 to meet funders/partners and advance biosecurity programs, fellowship development, and AI governance/evaluation collaborations.
APEC, Bangkok 2026
Can an AI govern its behaviour according to the characteristics of the situation in which an action is requested of it?
A pilot curriculum of six Socratic seminars for educators and other nontechnical learners.
Develops the Claim–Evidence–Independence framework to assess incident claims individually and proposes a reusable public evidence record for investigators and regulators.
Codes 121 top comments on a widely read essay about a July 2026 AI incident, finding commenters who adopted its anthropomorphic framing rarely named any possible action.
Testing whether AI agents can comply with corrective instructions after their earlier actions have already reduced or eliminated effective human control.
Travel support for an independent researcher from Bangladesh to present his single-author NeurIPS 2026 paper on how efficiently LLMs reason.
A model-agnostic method and tool for independently testing whether advanced AI systems remain stable and appropriately corrigible without requiring access to model internals.
Independently replicate and open-source a model-free AI agent monitor that caught 209 of 210 sabotage runs, plus static checks that find fail-open defects in safety tooling.
The Universal Stability Criterion (USG) project establishes a domain‑agnostic protocol that anticipates collapse in symbolic complex systems by detecting structural deviation before failure.…
The GRACE (Governed Release Architecture for Critical Environments) protocol establishes a six‑layer governance matrix that embeds stability and safety checks directly into the release pipeline of…
PGVP is a simple diagnostic protocol that reliably distinguishes true periodic generalization from spurious curve fitting, ensuring models are structurally validated before deployment.
Introduces Governance Drift and a reproducible framework for testing whether AI models maintain governance-aware behavior under user pressure across jurisdictions and domains.
Develops a behavioral comparison protocol using independently generated solutions and multi-round interrogation to detect injected backdoors that evade weak-model code monitoring.
Tests whether Qwen2.5-7B-Instruct internally represents Colombian identity, socioeconomic status, and stereotypes from Spanish and English prompts using layer-20 activation verbalization.
Tests whether sentence embeddings can recover a poisoned dataset’s hidden principal without clean data or a model, and maps the limits of this defense.
A non-LLM graph neural network compares code property graphs with trusted references to detect backdoors at low false-positive rates and provide node-level audit trails.
SpecCheck detects intent drift in LLM-formalized requirements using ambiguity elicitation and cross-model validation, evaluated on Lean 4 theorems and Python property-based tests.
The project evaluates secret-loyalty detections with matched controls and finds prior signals spurious, while verifying that organism C is byte-identical to the base model.
Develops an internal-representation scanning tool to detect hidden loyalties in AI models by comparing entity representations, and critiques existing sleeper-agent benchmarks.
