What Greg Brockman said, newest first. Changes are found by AI and checked before they show; every quote opens at its moment.
2026-09-14 · Bloomberg Podcasts · 06:48
“I think that our realization is that we need to pull back Earlier into our development process and training monitoring. There's always going to be a phase at which you do alignment, but you need to think about alignment as a core part of even this, this earlier phase.”He believes alignment must be treated as a core part of early development, not just deployment.
2026-09-14 · CrowdStrike · 1:11:40
“We always knew that these were going to become these important bottlenecks to development and not just be about deployment time but how we actually develop.”He says safety and security have become bottlenecks to AI development itself, not just deployment.
2026-09-14 · CrowdStrike · 1:24:15
“we're really seeing how much safety, security, and alignment are now becoming the bottleneck, right? The hardest part of what we have to do.”He says safety, security, and alignment are becoming the hardest part of what OpenAI has to do.
2026-09-14 · a16z · 03:20
“as we move to more capable models, you really have to make sure that safety, security, alignment, those are all standards that you're constantly upleveling.”Safety, security and alignment standards must be constantly uplevelled as models become more capable.
2026-09-14 · a16z · 40:23
“people shouldn't have to extract from the AI what it's capable of it should go the other way around the AI should say hey I can help you in this new way”AI should proactively tell people how it can help them rather than making them discover capabilities.
2026-09-12 · SkillCurb · 00:12
“AGI is a gradual thing, right? It's not really like a moment in time”He believes AGI arrives gradually rather than as a single moment in time.
2026-09-12 · SkillCurb · 07:43
“if we go forward two years and we look back, I think that that we will say this is when AGI was created”He expects that in two years people will look back and identify this period as when AGI was created.
2026-09-12 · SkillCurb · 06:15
“let's use that extra juice to increase the alignment properties, to increase the safety properties. Let's make sure that we have models that monitor other models in a trustworthy fashion”He believes the efficiency gains from AI should be used to improve alignment and safety, including models monitoring other models.
2026-09-12 · SkillCurb · 14:41
“I think that the amplification we've seen in software engineering we expect across every field for sure”He believes the amplification seen in software engineering will spread to every field.
2026-09-12 · SkillCurb · 08:47
“Who decides what's good and what's not good? Like, that is something that should be uniquely human”He believes judgment and taste about what is good should remain uniquely human.
2026-09-08 · TBPN · 07:51
“We're developing this safely. That's that's one of the core commitments that that we make and how we think about this.”He says OpenAI is developing these capabilities safely as a core commitment.
2026-09-08 · TBPN · 21:17
“I would look at all these things as timing and all about iterative deployment.”He views iterative deployment as the right approach even when early products fall below threshold.
2026-09-02 · TIME · 24:38
“as models get more capable you have to make sure that you increase your safety and security requirements in concert and that's exactly what we're doing”He believes safety and security must scale in lock step with model capability rather than being traded off.
2026-09-02 · TIME · 21:16
“if you make models extremely capable, you really want to make sure that as the capability grows that their alignment grows with them”He believes alignment and capability must grow together as models become extremely capable.
2026-08-17 · CNBC Television · 15:08
“That's why defenders have an upper hand, have an advantage, but need to lean in.”He believes defenders can win in cyber security if they act now and use AI tools to find and fix vulnerabilities before attackers.
2026-08-17 · CNBC Television · 27:37
“You can't separate out capability from safety, security, alignment. These are one thing, one system, one way.”He holds that capability cannot be separated from safety, security, and alignment.
2026-08-17 · CNBC Television · 04:58
“Because they're capable, it means that we need to ensure that they have alignment during training, that they need to be monitored during training in a robust way”He believes capable frontier models must be aligned and robustly monitored during training.
2026-08-17 · CNBC Television · 05:32
“I think that alignment is a non-negotiable.”He considers alignment during the training process non-negotiable.
2026-07-29 · Joanna Stern · 06:24
“And it really shifts towards a world where everyone becomes a manager, right? That everyone really is able to have all this leverage and empowerment.”He believes AI agents will turn everyone into a manager delegating tasks.
2026-07-29 · Joanna Stern · 16:34
“That what we need to have happen is far more compute needs to go into defending than the than attackers could bring to bear.”He believes more compute should go to defenders than attackers after the Hugging Face incident.
2026-07-14 · Zenvora Productions · 59:11
“I don't think that there's a sustainable state where the people who are building this technology and having successful products are not also investing super hard in safety”He argues safety is a core product feature and labs with successful products must invest heavily in it.
2026-06-27 · MindVoice Production · 14:45
“what we're seeing with systems we build at Open AAI, and they're still in early proof of concept stage, is that you are able to learn human preferences. you're able to learn what humans want from data”He believes human preferences can be learned from data, giving hope for solving value alignment.
2026-04-30 · Sequoia Capital · 16:21
“human attention is going to be this incredibly scarce resource, right? The doing of things now is easy.”He expects human attention to become the scarcest resource as AI does the doing.
2026-04-27 · Big Technology · 01:57
“Every piece of what we do needs to connect to the question of how do we continue to make progress, but also how do we make these models broadly available?”He believes every part of OpenAI's work must connect to making progress while making models broadly available.
2026-04-23 · Alex Kantrowitz · 20:35
“either extreme, reflexively, I don't think will yield the best outcome for the world.”He believes neither fully open nor fully restricted deployment of powerful models is the right approach, and that tactics should follow technical specifics.
2026-04-22 · Core Memory Podcast · 10:12
“you're not going to solve AGI going well for the world in a paper. It has to be a worldwide effort from contributions from society, from many different people”He believes ensuring AGI goes well for the world requires a broad societal effort rather than any single technical intervention.
2026-04-01 · Alex Kantrowitz · 33:50
“I think the reward is worth the risk, but I think that that is too too coarse grained of a of a answer in some sense.”He thinks the reward of AI development is worth the risk, but that framing is too coarse-grained.
Blumify can make mistakes. Please check important info.