Back
Apollo Research - AI Safety Evaluation Organization
webCredibility Rating
4/5
High(4)High quality. Established institution or organization with editorial oversight and accountability.
Rating inherited from publication venue: Apollo Research
Apollo Research is a key third-party evaluator in the AI safety ecosystem, providing independent assessments of frontier models for dangerous capabilities and advising policymakers; their work on scheming evaluations is directly relevant to deceptive alignment concerns.
Metadata
Importance: 62/100homepage
Summary
Apollo Research is an AI safety organization focused on evaluating frontier AI systems for dangerous capabilities, particularly 'scheming' behaviors where advanced AI covertly pursues misaligned objectives. They conduct LLM agent evaluations for strategic deception, evaluation awareness, and scheming, while also advising governments on AI governance frameworks.
Key Points
- •Specializes in evaluating frontier AI for 'scheming' - covert pursuit of misaligned objectives by advanced AI systems
- •Conducts LLM agent evaluations focused on strategic deception, evaluation awareness, and scheming detection
- •Partners with major AI labs including OpenAI, Google DeepMind, Microsoft, and Amazon
- •Supports governments and international organizations in developing technical AI governance regimes and evaluation standards
- •Conducts fundamental research into emergence of scheming behaviors and potential mitigations
Cited by 15 pages
| Page | Type | Quality |
|---|---|---|
| Large Language Models | Concept | 62.0 |
| AI Risk Cascade Pathways Model | Analysis | 67.0 |
| AI Risk Warning Signs Model | Analysis | 70.0 |
| Apollo Research | Organization | 58.0 |
| Survival and Flourishing Fund | Organization | 59.0 |
| Capability Elicitation | Approach | 91.0 |
| AI Governance Coordination Technologies | Approach | 91.0 |
| Dangerous Capability Evaluations | Approach | 64.0 |
| Evals-Based Deployment Gates | Approach | 66.0 |
| AI Evaluations | Research Area | 72.0 |
| AI Evaluation | Approach | 72.0 |
| Third-Party Model Auditing | Approach | 64.0 |
| Sleeper Agent Detection | Approach | 66.0 |
| Technical AI Safety Research | Crux | 66.0 |
| AI Development Racing Dynamics | Risk | 72.0 |
Cached Content Preview
HTTP 200Fetched Mar 20, 20263 KB
# Dedicated to improving our understanding of AI to mitigate its risks. Our mission ### AI systems will soon be integrated into large parts of the economy and our personal lives. While this transformation may unlock substantial personal and societal benefits, there are also vast risks. We think some of the greatest risks stem from “scheming” AIs, i.e. advanced AI systems that covertly pursue misaligned objectives. Our goal is to understand and evaluate for the emergence of scheming well enough to prevent the possible harms that scheming AIs might cause. About us ### Apollo Research is focused on reducing risks from dangerous capabilities in advanced AI systems, especially scheming behaviors. We design AI model evaluations and conduct technical research to better understand state-of-the-art AI models. Our governance team provides global policymakers with expert technical guidance. What we do [Our research](https://www.apolloresearch.ai/research/) We develop and run evaluations of frontier AI systems. Our expertise is in LLM agent evaluations for strategic deception, evaluation awareness and scheming. We also conduct fundamental research into the emergence of scheming and into potential mitigations. We support governments and international organisations by developing technical AI governance regimes. Expertise in building a robust third-party evaluation ecosystem, effectively regulating frontier AI systems, and establishing standards and best practices. Additionally, we provide consultancy services for building responsible AI development frameworks, designing research programs, ecosystem mapping and literature reviews, and more. Our partners Frontier labs, multinational companies, governments, and foundations partner with Apollo Research.         “One thing you might imagine is testing for deception for example, as a capability. You really don’t want that in the system because then you can’t rely on anything else it’s reporting. So that would be my number one emerging capability I think that would be good to test for.” **Demis Hassabis** CEO Google DeepMind Contact [Get in touch](https://www.apolloresearch.ai/contact/) ## For collaborations and other inquiries, please get in touch Currently, we are looking for collaborators in the broader AI governance, policy, and strategy sphere, and for partnerships with leading AI developers for model evaluations. [Get in touch](https://www.apolloresearch.ai/contact/)
Resource ID:
329d8c2e2532be3d | Stable ID: NDNkMjgwMT