our story
About Apollo
Our goal is to secure frontier AI systems across development, deployment, and governance.
.webp)
What we do
Our mission
AI systems will soon be deeply integrated into the global economy and everyday life. While this brings personal and societal benefits, it also introduces serious risks. We believe some of the most dangerous risks come from scheming frontier AI: advanced AI systems covertly pursuing misaligned objectives.
We believe that robustly mitigating scheming risks will require a deep scientific understanding of the emergence of scheming. Our goal is to secure frontier AI systems across development, deployment, and governance.
Our journey
2023
First movers on frontier risks
We identified scheming AI as a serious risk early and built our research around it. We:
presented at the inaugural AI Safety Summit in the UK.
shared policy recommendations with the UK’s Frontier AI Taskforce.
established ourselves as leaders in AI evaluations
Our work was covered by the BBC, Bloomberg, and presented before the US Senate.
.webp)
.webp)
2024
Detecting scheming at the frontier
Scheming moves from theory to evidence. We:
partnered with OpenAI to test their o1 model before public deployment
published the first evidence that frontier models can scheme in context
engaged with policymakers at the EU AI Office, UN Advisory Body, and US Congress.
Our findings shifted the field and were covered by TIME and MIT Technology Review. Our research entered into the US Congressional record.
.webp)
2025
Shaping research agendas
We moved from identifying risks to shaping how the world addresses them. We ran evaluations for all major labs as well as:
partnered with OpenAI to study anti-scheming interventions on frontier models
published the first analysis of internal AI deployment risks
created a loss of control playbook to help governments prepare for advanced AI threats.
Our work was featured in NY Times, Nature, and The Economist.
.webp)
.webp)
.webp)
2026
Scaling for what’s ahead
We are now a Public Benefit Corporation to achieve our mission of secure AI development and deployment, and have opened an office in San Francisco. We are:
building a science of scheming to predict and prevent scheming risks
advising governments and developing policy recommendations on loss of control threats
building an AI security tool to monitor frontier AI agents.
Join our team.
Our blog
Updates from Apollo
Towards embedded evaluations for scheming propensities
Embedded evaluations for frontier AI scheming: four claims every safety case must make, and the persistent, deep access embedded evaluators need to verify them.
On Testifying on Misaligned AI in the U.S. Senate
Apollo Research CEO Marius Hobbhahn testifies before the U.S. Senate on AI scheming: why our tools for detecting misaligned AI are degrading, and four steps to act before models learn to hide deception.
Principles for Embedded Evaluations
A framework for embedded evaluations of frontier AI developers, centered on claim-based, publicly reported assessments.
watcher
Your AI agent monitoring tool by Apollo Research
Watcher catches dangerous coding agent behavior before it becomes an incident.
