The Alignment Problem Isn’t Theoretical
Deception, Deployment, and automated testing with PETRI
19 min
Writing
Essays on AI safety, biosecurity, and methodology
Long-form writing by Austin P. Morrissey on AI capability evaluations, biosecurity, and the philosophy of empirical science.
These essays cross-post to my Substack, where they reach a wider audience. The canonical version lives here, and Substack versions point back via canonical link.
Featured
No matching items
Archive
What makes for good research managers, mentors, and leaders?
research
leadership
mentorship
Evaluating LLM hypothesis generation in biology is hard
ai-evals
biosecurity
scientific-method
MIT Lincoln Laboratory & The Dual-Use Frontier
biosecurity
dual-use
science-policy
mit
Making Software for Your Non-Tech Small Business
ai-tools
software
productivity
Biorisk & DNA fragment analysis with Claude
biosecurity
dual-use
ai-safety
In Good Company, a Letter to [Redacted]
political
voice
dialogue
Networking for Spies: Translating a Cyrillic Text with Claude Code
ai-tools
claude-code
applied-ai
Autonomous Web Browsing with Comet
ai-safety
agents
security
The Alignment Problem Isn’t Theoretical
ai-safety
alignment
evals
Keeping our Teeth in Biotech; prelude
biotech
philosophy
vision
No matching items