Field: Jailbreaking AI models requires conversational questioning rather than fancy JSON
Dylan Field · Dylan Field on the “Permanent Underclass of Zero Taste” · Sourcery with Molly O'Shea · Jul 2, 2026 · at 31:04
Figma CEO Dylan Field demystifies how AI safety guardrails get bypassed during natural interactions rather than complex exploits.
0:00 / 0:09exact quote · 9.0s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“It's not like you have to do some fancy JSON thing. Like, that's what people like to popularize it online with, but I mean, you can just talk to them and ask them questions.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Dylan Field
Field: AI judgment and design ability do not scale with IQ points
“All my experience with AI so far have shown that on formal domains, things that are verifiable you can get to amazing outcomes with enough IQ points, and yet judgment is like somehow not purely correlated with just adding IQ points to the model.”
Field: Figma engineers multiply productivity tenfold using specialized AI agent swarms
“We have plenty of people at Figma that are and are getting great results though.
And it's amazing how much we can do there when you have folks that really understand the infrastructure that they're trying to build.
And they go and you know, they're already 10 …”
Field: Dismissing team roles due to AI is a temporary hype phase
“And I think that we're in a temporary phase of people really, you know, getting excited about what they can do with AI. But not yet realizing all the limitations that they have”
Field: Design AI evaluations are inherently non-verifiable and require human judgment
“And the evals we're running on stuff we're doing, stuff others are doing, like a lot of it is inherently non-verifiable. It's like, gotta be human judged at the end of the day. And you can set up more automatic methods for that, but you still have that human i…”
Field: Major AI companies will not become monolithic value-sucking black holes
“There's going to kind of be like a black hole and suck it all up. Which I'm sure that that's not going to be what ends up happening. I think that they're all going to get a lot smarter than that. And I think most don't even want to do that in the first place”