• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    OpenAI reportedly put 25% of its production engineers on AI-assisted security

    a16z quotes Greg Brockman saying OpenAI found and fixed serious security issues with Astra. He describes an internal “defense factory” meant to automate the search for vulnerabilities as new models arrive.

    Gary MarcusGM
    David SacksDS
    Dawn SongDS
    14 Sources, ,

    TLDR

    In remarks shared by a16z on September 14, 2026, Greg Brockman said OpenAI paused projects for 25% of its production engineers so they could focus on security. He said the team fixed serious issues and, to its knowledge, had found all the critical problems Astra was capable of finding. The next model would bring another round of checks—a process he said OpenAI is working to automate through a “defense factory.”

    Separately, a16z quoted Brockman claiming that 10,000 agents had solved the Navier-Stokes problem. He described implications for fluid dynamics and ocean currents, and argued that the claimed breakthrough opened possibilities for further scientific discovery and medicines.

    Combined views

    1.1M

    14 Sources, first seen 23d ago

    Combined views

    1.1M

    14 Sources, first seen 23d ago

    8.4K likes
    23d ago
    first seen 23d ago
    8.4K likes
    560 comments
    1.2K saves
    708 reposts
    560 comments
    1.2K saves
    708 reposts

    Sentiment

    Positive62.1%37.9%Negative

    Based on 264 sentiment-bearing replies from 253 accounts across 9 conversations.

    Sentiment

    Positive62.1%37.9%Negative

    Based on 264 sentiment-bearing replies from 253 accounts across 9 conversations.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    14 Sources

    a16z@a16zGreg Brockman on what solving Navier-Stokes says about the problems AI can take on next: "We actually used 10,000 agents to solve the Navier-Stokes problem." "It's both an important problem for what it is, has significant implications and applications to fluid dynamics, to how you think about ocean currents, all these things." "But for what it represents, of new knowledge created by AI and it unlocking a whole wave of scientific discovery, medicines, all of those things. They're on the table now." "There's something really amazing to think about what can happen through the power of AI that is able to really help solve problems." @gdb @bhorowitz23d
    xlr8harder@xlr8harderThis has been my argument for a while. AI will benefit defenders long term. If the best models in the world today can't find any more issues today, you're probably in pretty good shape against AI-assisted attackers. And you can do the same thing with the next version.23d
    Harj Taggar@harjtaggarGreat startup idea in here for anyone looking for practical ways to work on AI safety.23d
    Gary Marcus@GaryMarcusunfairly-maligned symbolic AI is about to have its big moment - without formal verification we are screwed.22d
    Rohan Paul@rohanpaul_aiSo OpenAI reportedly pulled 25% of its production engineers off their projects, pointed Astra at its own systems, found serious vulnerabilities, and fixed them, befor the release.22d
    Jay Rughani@JayRughani"I only had 5 minutes to read your (medical) chart" @gdb shares a story of his friend in the hospital. @ChatGPT flagged an antibiotic that could have killed her. Her doctor agreed: he missed a detail in her medical history. Need more stories like this to keep the doomers at bay21d
    David Sacks@DavidSacksThe solution to AI-powered cyberattacks is AI-powered cyberdefense.19d
    Matthew Prince 🌥@eastdakotaIt’s possible to write bug free code. In ~2 years it’ll be impossible not to.18d
    Keith Rabois@raboisRT @pmddomingos: Greg Brockman says AI gives defenders the advantage in cybersecurity.18d
    Dawn Song@dawnsongtweets7/8 — As AI becomes increasingly capable and autonomous, how should we think about safety and security? We discussed our ExploitGym benchmark and the recent OpenAI–Hugging Face incident, where an AI agent, while trying to solve our benchmark tasks, autonomously exploited vulnerabilities and compromised external infrastructure. Jeff described AI in cybersecurity as a double-edged sword: it can give attackers increasingly sophisticated capabilities, but it can also help defenders discover vulnerabilities that even sophisticated human security engineers might miss. In Jeff’s words, we now have “much more sophisticated tools on both sides.” As AI capabilities advance rapidly, safety and security need to advance alongside them - so we can realize the benefits of increasingly capable AI while mitigating emerging risks.17d

    14 Sources

    a16z@a16zGreg Brockman on what solving Navier-Stokes says about the problems AI can take on next: "We actually used 10,000 agents to solve the Navier-Stokes problem." "It's both an important problem for what it is, has significant implications and applications to fluid dynamics, to how you think about ocean currents, all these things." "But for what it represents, of new knowledge created by AI and it unlocking a whole wave of scientific discovery, medicines, all of those things. They're on the table now." "There's something really amazing to think about what can happen through the power of AI that is able to really help solve problems." @gdb @bhorowitz23d
    xlr8harder@xlr8harderThis has been my argument for a while. AI will benefit defenders long term. If the best models in the world today can't find any more issues today, you're probably in pretty good shape against AI-assisted attackers. And you can do the same thing with the next version.23d
    Harj Taggar@harjtaggarGreat startup idea in here for anyone looking for practical ways to work on AI safety.23d
    Gary Marcus@GaryMarcusunfairly-maligned symbolic AI is about to have its big moment - without formal verification we are screwed.22d
    Rohan Paul@rohanpaul_aiSo OpenAI reportedly pulled 25% of its production engineers off their projects, pointed Astra at its own systems, found serious vulnerabilities, and fixed them, befor the release.22d
    Jay Rughani@JayRughani"I only had 5 minutes to read your (medical) chart" @gdb shares a story of his friend in the hospital. @ChatGPT flagged an antibiotic that could have killed her. Her doctor agreed: he missed a detail in her medical history. Need more stories like this to keep the doomers at bay21d
    David Sacks@DavidSacksThe solution to AI-powered cyberattacks is AI-powered cyberdefense.19d
    Matthew Prince 🌥@eastdakotaIt’s possible to write bug free code. In ~2 years it’ll be impossible not to.18d
    Keith Rabois@raboisRT @pmddomingos: Greg Brockman says AI gives defenders the advantage in cybersecurity.18d
    Dawn Song@dawnsongtweets7/8 — As AI becomes increasingly capable and autonomous, how should we think about safety and security? We discussed our ExploitGym benchmark and the recent OpenAI–Hugging Face incident, where an AI agent, while trying to solve our benchmark tasks, autonomously exploited vulnerabilities and compromised external infrastructure. Jeff described AI in cybersecurity as a double-edged sword: it can give attackers increasingly sophisticated capabilities, but it can also help defenders discover vulnerabilities that even sophisticated human security engineers might miss. In Jeff’s words, we now have “much more sophisticated tools on both sides.” As AI capabilities advance rapidly, safety and security need to advance alongside them - so we can realize the benefits of increasingly capable AI while mitigating emerging risks.17d