The Vergecast · Tuesday, August 4, 2026
David Pierce questions whether proponents of open weight AI models are actively trying to dismantle safety or genuinely believe openness is the path to it. Robert Hart suggests it's likely a mix, noting that while open models have benefits like broad accessibility and collaborative problem-solving, they also grant access to bad actors.
“And so maybe it's not possible to have AI that is both perfectly safe and perfectly open.”
“And that's kind of the the crux of the debate we're having.”
“I think that the industry, or at least the people who are working within the industry, are much more inclined to say, we're going to prioritize open source, we're going to prioritize accessibility, we're going to make sure that this technology gets out to as many people as possible.”
“And the counterpoint is that those are the very same forces that are going to lead to some incredibly dangerous outcomes.”
“This is a bit of a crazy thing to ask, but do you think that the industry, or at least the people who are arguing for open weights, are actively trying to dismantle safety.”
“Or do you think that they just genuinely believe that the best way to achieve safety is through openness?”
“I think that's a really hard question to answer.”
“I think there are probably elements of both.”
“The counterpoint to that is that it also means that bad actors will have access to the same tools and potentially be able to use them for nefarious purposes.”