Last Week in AI · Tuesday, August 11, 2026
The podcast discusses the significant risks associated with open-source AI models in the context of biosafety, with one speaker arguing that there is no counter-argument to the idea that open-source models capable of bio-weapon design could greatly increase the destructive capabilities of malicious actors. They note that even if models have safeguards, they can be disabled or have loopholes exploited.
“I think it's definitely a fiction when it comes to bio. I have yet to hear a single person articulate an argument that makes open source bio capable, bio weapon design capable models that can materially increase the essentially destructive footprint of any psycho terrorist group, nation state proxy who wants to launch a bio weapon dramatically.”
“I think the vulnerability comes from the models themselves. And yes, they might have safeguards, but they could have been disabled. And if they weren't disabled, there might be loopholes that are now being leveraged by these different actors.”
“And if you don't have the ability to patch those models or to get them to update their safeguards, that's a problem.”