Transcriptly

Why Every Founder Needs to Care About AI Safety Right Now | Glen Wise, Cinder

EO09:43Added Aug 31, 2026

Glen Wise, co-founder and CEO of Cinder, went from hunting terrorists for the US Government to fighting AI's darkest threats—raising $14M in under a year. While Silicon Valley raced to "move fast and break things," Glen and his team saw what others missed: platforms had zero infrastructure to stop deepfakes, nation-state attacks, and non-consensual intimate imagery at scale.

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:00The Pandora box of technology is already
  3. 00:02open. The reality is that kids are
  4. 00:03growing up with AI and they're talking
  5. 00:05to systems that are saying yes, that are
  6. 00:07agreeing to them, that are placating
  7. 00:09them because that's what these systems
  8. 00:10are trained to do. At the same time,
  9. 00:11it's being used for really horrible
  10. 00:14things. I think there's a real there's a
  11. 00:16real worry. Everything from just spam
  12. 00:18and AI slop from that to highly
  13. 00:21targeted, highly contextual fishing
  14. 00:23campaigns to victimizing people by
  15. 00:26creating non-consensual intimate
  16. 00:28imagery. Someone needs to be able to
  17. 00:30stop those horrible things from
  18. 00:31happening so innovation can continue
  19. 00:34because if you don't what's going to end
  20. 00:35up happening is that there's going to be
  21. 00:36regulation. They're going to have to
  22. 00:38stop and fix things and before they can
  23. 00:39innovate and it slows down ability to
  24. 00:42innovate as a society. And so what I
  25. 00:43want to do is be able to we should be
  26. 00:45able to be the ones that are stopping
  27. 00:47those horrible things and innovating on
  28. 00:48that and allow everyone else to innovate
  29. 00:51incredibly quickly. I'm Glenn Wise. I'm
  30. 00:54the co-founder CEO of Cinder, a central
  31. 00:56platform for trust and safety
  32. 00:57operations. You can think of Cinder as
  33. 01:00the club bouncer for the internet. Our
  34. 01:03customers want to be able to set rules
  35. 01:06on what they allow on their platforms
  36. 01:07and what they don't allow. And so Cinder
  37. 01:09is the system that they use to enforce
  38. 01:12those rules across all of their users.
  39. 01:15We raised $14 million from investors
  40. 01:17such as Y Cominator and Excel. Our
  41. 01:19mission is to make the internet a safer
  42. 01:21place.

  43. A Greater Mission to Safeguard America

  44. 01:31So growing up in New Jersey, being 6 or
  45. 01:347 years old during 911, that was such an
  46. 01:36impactful sort of moment in my life and
  47. 01:39my community. And when you grow up with
  48. 01:41that in the backdrop, any opportunity to
  49. 01:44be part of a greater mission was really
  50. 01:46important to me. So then when I went to
  51. 01:48college and I realized that you can
  52. 01:50actually go work at the NSA or the CIA
  53. 01:53or the FBI on these exact issues of
  54. 01:56making sure that never happens again.
  55. 01:57That's the the only thing I wanted to
  56. 01:59do. And so that inevitably led me to
  57. 02:01join government being able to work on
  58. 02:02counterterrorism issues which inevitably
  59. 02:04led to me doing the same type of work at
  60. 02:06at Facebook on a threat intelligence
  61. 02:08team. Facebook started investing heavily
  62. 02:10in threat intelligence before the harm
  63. 02:13became prevalent on platform. Being on a
  64. 02:16red team, I think, is one of the most
  65. 02:17fun jobs that someone could have and
  66. 02:20someone can do because you get to, you
  67. 02:22know, try to break things. This is
  68. 02:23something that I instill on my team
  69. 02:24today. It's like, how do you have that
  70. 02:25adversarial mindset, right? There's
  71. 02:27going to be people on the other side of
  72. 02:28that trying to stop you and you have to
  73. 02:29acknowledge that and be able to think
  74. 02:31like that. At the time, I was talking to
  75. 02:33so many other companies that were really
  76. 02:35struggling with getting a grasp on the
  77. 02:38abuse occurring on their platform. And
  78. 02:39so, I thought that this would be a
  79. 02:40really interesting thing to start. I was
  80. 02:43thinking about it for a long time to the
  81. 02:44point where, you know, it's all I
  82. 02:46thought about outside of work. Part of
  83. 02:48my hesitation of founding Cinder that I
  84. 02:51just thought I wasn't ready, right? I
  85. 02:53felt I didn't have the right experience.
  86. 02:54My initial thought at the time was to go
  87. 02:58to a smaller company and help them
  88. 02:59build, you know, a platform internally.
  89. 03:01So, I ended up just reaching out to a
  90. 03:03bunch of people. And I realized that
  91. 03:05there were a lot of companies to choose
  92. 03:08from to build something internally that
  93. 03:11at that point you should just build it
  94. 03:13once centrally and then give it to all
  95. 03:16these companies. All right. It just like
  96. 03:17made so much sense as a business. That's
  97. 03:19inevitably why I made the decision to to
  98. 03:22found it. And I think that it's
  99. 03:23important. Someone needs to be able to
  100. 03:25stop those horrible things from
  101. 03:26happening so innovation can continue.
  102. 03:29And so what I want to do is be able to
  103. 03:30we should be able to be the ones that
  104. 03:32are stopping those horrible things and
  105. 03:34innovating on that allow everyone else
  106. 03:36to innovate incredibly quickly and
  107. 03:38deploying new things and building new
  108. 03:40things without worrying about this
  109. 03:42underbelly of the internet and these
  110. 03:44underbelly of of harmful people. I

  111. How I Got Into YC

  112. 03:47applied to Y Comer thinking that nothing
  113. 03:49would happen, you know, literally an
  114. 03:51hour or two before the deadline was due.
  115. 03:52Uh ended up getting a call from Michael
  116. 03:55Cyel and Michael told me that he thought
  117. 03:57the idea was great. He's really excited
  118. 03:59about it. He wants us to he wants me to
  119. 04:01do it, but I need to find a co-founder.
  120. 04:04And I told him that there were people
  121. 04:06that I worked with at Facebook that I
  122. 04:07thought were really amazing people that
  123. 04:09I really wanted to work with, but it's
  124. 04:11going to be impossible for me to get
  125. 04:13them to to leave Facebook and to
  126. 04:15actually join starting this. And fast
  127. 04:17forward 3 months later, had a really
  128. 04:19great team of people that I was really
  129. 04:21respected. Um, and it was amazing that I
  130. 04:24mean still is amazing that I get to work
  131. 04:25with them every day. Bring all these
  132. 04:27people on and and work with them was a
  133. 04:30really special moment for me.

  134. Solving Problems That Impact Society

  135. 04:36This year alone with Cinder, our
  136. 04:38customers have submitted over 100,000
  137. 04:40reports. The National Center for Missing
  138. 04:42and Exploited Children. That alone is a
  139. 04:44pretty insane number, but it's also just
  140. 04:46like scratching the surface of all of
  141. 04:48the really bad things. We see abuse
  142. 04:50today that 4 years ago only a nation
  143. 04:54state could do. And today, a teenager
  144. 04:56with a laptop is able to do it. As a
  145. 04:58society, we need systems that can combat
  146. 05:00this abuse before it turns into a total
  147. 05:05hellscape. When we set out to build
  148. 05:06Cinder, we knew that we were going to
  149. 05:08build a platform and that is really
  150. 05:10challenging and I would encourage anyone
  151. 05:13to not build a platform from the
  152. 05:15beginning because you have to build so
  153. 05:17many things on top of it in order for it
  154. 05:19just to work, right? You have to be able
  155. 05:21to integrate all these signals. You have
  156. 05:23to be able to make it highly
  157. 05:24configurable and customizable. You have
  158. 05:26to be able for customers to represent
  159. 05:28their own unique data and their own
  160. 05:30unique workflow within it. So we did a
  161. 05:32lot of work initially to support all
  162. 05:36these different content types to support
  163. 05:37all these different types of workflows
  164. 05:39that takes a long time. We set out to
  165. 05:41build what was effectively a threat
  166. 05:43intel platform for trust and safety. And
  167. 05:45we thought this was a great idea because
  168. 05:47we saw that pain in previous jobs and we
  169. 05:50were going to go solve it and bring it
  170. 05:52to everyone else. So we spent all this
  171. 05:53time building this platform and then we
  172. 05:56went out to show people and they
  173. 05:58effectively told us that this was
  174. 06:00interesting but they did not want it and
  175. 06:02what they really needed was a way to
  176. 06:04handle their overall operations and so
  177. 06:07we invested all this time uh in me and
  178. 06:10Declan one of the co-founders spent so
  179. 06:12much time building this thread intel
  180. 06:13platform to be told that they wanted
  181. 06:16something else and that's like a huge
  182. 06:17learning for us right and that that
  183. 06:19brings us to this idea of really
  184. 06:22focusing ing on what the core customer
  185. 06:24problem is and get that problem down
  186. 06:27really tight and then go out and build.

  187. The Dark Side of AI Innovation

  188. 06:33And my focus from like the beginning of
  189. 06:35my career has been on how do you stop
  190. 06:37people from taking advantage and and
  191. 06:39abusing other systems or or people. And
  192. 06:42so being able to empathize with the
  193. 06:43customer was really key, right? Even
  194. 06:45when we had nothing built, someone being
  195. 06:47able to listen to them and understand
  196. 06:50what they were going through and provide
  197. 06:52suggestions on how to fix it and then a
  198. 06:55few weeks later come with a product. It
  199. 06:57solved a real problem. So a really
  200. 06:59prominent harm facing Gai customers
  201. 07:02today is basically non-consensual
  202. 07:04intimate imagery NCI. We are seeing
  203. 07:06harms become more and more complex as
  204. 07:10time goes on. At Cinder, we have a
  205. 07:12framework called the decision spectrum.
  206. 07:14And almost every problem can be placed
  207. 07:16on the spectrum in some way where on the
  208. 07:18left side of the spectrum is a really
  209. 07:19something really simple, like a really
  210. 07:20simple decision of yeah, this image is
  211. 07:22nudity or this uh chat message is fraud,
  212. 07:27right? Like a really simple
  213. 07:28identification of a single piece of
  214. 07:30content or a single event. And on the
  215. 07:31other side, you're investigating some,
  216. 07:34you know, a really complex nation state
  217. 07:37or really complex network or an
  218. 07:40adversary that is advanced in trying to
  219. 07:42hide their behavior. There are way too
  220. 07:44many things to review and prioritizing
  221. 07:47is really difficult because everything
  222. 07:50is a heart, right? Everything is hurting
  223. 07:53someone else. And that burnout is real.
  224. 07:56and you see people burning out the
  225. 07:57industry all the time because they
  226. 08:00almost feel like there's they're facing
  227. 08:02an uphill battle of targeting this. So I
  228. 08:04think our team, you know, we we
  229. 08:07definitely see some of that, but for us,
  230. 08:10everyone at our company is deeply
  231. 08:12motivated by the mission and everyone
  232. 08:14wants to inevitably have a world where
  233. 08:18the internet and innovation is is safe
  234. 08:21and can be used by anyone. And our
  235. 08:23founding team had direct experience in
  236. 08:25building these systems at many different
  237. 08:28places. And we experienced that harm
  238. 08:31firsthand, right? We experienced that
  239. 08:33painoint firsthand. I think in this
  240. 08:36space, the only way that you're able to
  241. 08:39build an effective product is by having
  242. 08:40that experience, but also having that
  243. 08:43attraction to the mission and having
  244. 08:45that attraction to the outcome and
  245. 08:47having that purpose behind it all
  246. 08:49because the the work is so difficult. It
  247. 08:52is very complex and very nuanced that
  248. 08:54you have to have that that drive. When
  249. 08:56you hear stories from our teams of
  250. 08:58something that we took down or something
  251. 08:59that we helped stop or an abuse that we
  252. 09:02caught that allows us to go on, right,
  253. 09:05and be really motive for the next day.
  254. 09:07And so I think you can it's you you can
  255. 09:08easily get bogged down in the harm,
  256. 09:11right? You can easily look at all of the
  257. 09:13really horrible things happening across
  258. 09:14the internet and and be really dejected
  259. 09:17from it, but know that there's a path to
  260. 09:20being able to come at it. and someone
  261. 09:21needs to combat it. And if it's not us,
  262. 09:23then
  263. 09:25you know who else is that going to be?