OpenAI is backing a research project to try to reduce the odds of artificial intelligence making a new biological weapon.
On Monday the Sam Altman–led start-up said it had selected 14 organizations from more than 400 applications to receive grants aimed at ensuring advances in AI benefit everyone. Among them is the Nuclear Threat Initiative (NTI), a nonprofit dedicated to countering nuclear and biological catastrophe. The money will fund a project to better share information on potential AI biothreats across international borders.
The goal is to have a system where information on risks is easily accessible to international safety experts and researchers, says Hayley Severance, NTI’s deputy vice president for global biological policy and programs.
“It’s really to look at AI biorisk mitigations and ensuring that we’re working across borders to share information about those specific risks and how we might address them,” she says. “The issue isn’t [just] across [borders]; it’s also across labs.”
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
The legalities of using AI to design or alter organisms or compounds that can pose a biological risk is complex, Severance says. Different jurisdictions have different regulations, and that can cause confusion when it comes to sharing information between labs, she says.
“We’re basically looking into ‘What are the barriers and potential avenues to facilitate information sharing?’ so that, when people are coming to the table, either from model developers or biosecurity experts, to discuss these issues, they know the boundaries in which they can have those discussions,” she says.
The ultimate goal is better monitoring for the possible misuse of AI to make new biological weapons. Already, scientists have shown AI can design and synthesize a new virus not found in nature.
“Sophisticated actors who had traditionally been working in very high-end facilities with lots of resources are now able to do even more as AI improves our understanding of how to manipulate biology,” Severance says. “We’re worried about those who are doing sophisticated things being able to do even more sophisticated or targeting type attacks.”
It is not clear how much funding the NTI received from OpenAI, although the start-up said it had committed $1 million total to the new grants. Each awardee is also eligible to receive up to $1 million in API credits. NTI will use their tokens to help tag information to “expand the depth and scale of analysis,” Severance says.
“Through these new grants, researchers and practitioners will turn high-level ideas into concrete work, examining what could work, what it might cost and how approaches could be implemented responsibly,” OpenAI said in a statement.