Michael Chen
-
Michael works on evaluations-based AI governance at METR. Michael has advised various AI companies on their frontier safety policies, including Amazon, Google DeepMind, and G42. Prior to joining METR, he contributed to research studying AI deception and hazardous knowledge in large language models. His research has been covered in outlets such as The Guardian, MIT Technology Review, and Time. Michael previously worked on cloud infrastructure and AI applications at Stripe.
-
Frontier AI safety policies
Third-party oversight of frontier safety
Safety cases
International coordination
Loss of control evaluations and mitigations
EU AI Act Code of Practice
U.S. AI Action Plan implementation
U.S. state bills
Authoritarian misuse of AGI
Optimizing moral value of the long-term future
In addition to the above topics, I would also be interested in supervising additional topics, including technical projects:
Extending RE-Bench to evaluate AI R&D to increase an AI's dangerous capabilities in bio or cyber
Threat modeling AI takeover, either through internal or external deployments
Realistic agentic environments for AI development with opportunities for sabotage
Forecasts for AI x-risk conditional on evaluation results
Realistic agentic environments for honesty or law-following
I'm especially interested in research that could integrate into METR’s recommendations for AI company and government policy.
Examples of policy reports I've previously worked on:
-
ML knowledge: know basic algorithms (sgd, logistic regression, L1 regularization, …) and their properties; know the transformer architecture.
Enthusiasm about the subject!!
Ability to try things until they work
Decent writing skills
Decent software engineering skills
Good at using AI to make them better at writing and software
Member of Policy Staff, METR