Name: Could an AI model help create the next pandemic?
Start: 2025-11-11T17:45:00+13:00
End: 2025-11-11T19:45:00+13:00
Location: EPIC Innovation

As language models achieve startling performance in complex fields like virology, this question becomes critically important.

Join [Ruben Castaing](https://www.linkedin.com/in/ruben-castaing/?originalSubdomain=nz) from the University of Canterbury for a vital discussion on the intersection of AI and biosecurity. He will introduce techniques to precisely identify and remove dangerous biological weapons knowledge from within Large Language Models.

While addressing the current flaws and uncertainties of this method, Ruben will also explore how this work could be a key to solving other fundamental AI safety problems, including model deception and general misalignment.

**Links for further reading**
**Paper:** [Deleting Biological Weapons from Large Language Models](https://rubencastaing.github.io/Deleting-Biological-Weapons-Knowledge-from-LLMs/Deleting%20Biological%20Weapons%20Knowledge%20from%20Large%20Language%20Models%20(1).pdf)
**Less Wrong: [Creating an AI Safety Fellowship](https://www.lesswrong.com/posts/88cEgGcSXwFvPms4s/i-made-an-ai-safety-fellowship-what-i-wish-i-knew)**
**[RegulateAI.NZ](http://regulateai.nz) \- Sign the Petition**
**[Probably Private - How does Machine Unlearning Work](https://probablyprivate.com/issue/how-does-machine-unlearning-work/)**

emma.ai.strategy

Ruben Castaing

Christchurch AI Risk and Ethics Community

Technology

Risk Management

Risk Governance and Compliance

Artificial Intelligence

Machine Learning

Could an AI model help create the next pandemic? 

EPIC Innovation

Share

Christchurch AI Risk and Ethics Community

Could an AI model help create the next pandemic?

Christchurch AI Risk and Ethics Community

Details

Members are also interested in