Open AI locks down Astra
about 2 months ago
OpenAI said it would more tightly monitor new AI and moved development of its Astra model into isolated test environments after unplanned hacking attacks raised concerns it could execute cyberattacks.
Write your reply
by Grady Booch Reposted by Joanna Bryson
Water will find its way through cracks in walls and foundations in such a manner that even the most wondrously designed structure may collapse into pile of rubble.
Do not attribute agency to the water; rather, denounce the engineers and the builders and the maintainers who failed in their work.
Do not attribute agency to the water; rather, denounce the engineers and the builders and the maintainers who failed in their work.
OpenAI puts the brakes on a new model because it’s supposedly too powerful
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The…
🔗 https://hnow.live/a/d1289836
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The…
🔗 https://hnow.live/a/d1289836
OpenAI puts the brakes on a new model because it’s supposedly too powerful
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The announcement follows its recent disclosure that OpenAI models accidentally hacked Hugging Face…
hnow.live
August 7, 2026 at 7:33 PM
Water will find its way through cracks in walls and foundations in such a manner that even the most wondrously designed structure may collapse into pile of rubble.
Do not attribute agency to the water; rather, denounce the engineers and the builders and the maintainers who failed in their work.
Do not attribute agency to the water; rather, denounce the engineers and the builders and the maintainers who failed in their work.
Reposted by Grady Booch
OpenAI puts the brakes on a new model because it’s supposedly too powerful
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The…
🔗 https://hnow.live/a/d1289836
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The…
🔗 https://hnow.live/a/d1289836
OpenAI puts the brakes on a new model because it’s supposedly too powerful
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The announcement follows its recent disclosure that OpenAI models accidentally hacked Hugging Face…
hnow.live
August 7, 2026 at 7:01 PM
OpenAI puts the brakes on a new model because it’s supposedly too powerful
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The…
🔗 https://hnow.live/a/d1289836
OpenAI says it is pausing "internal activities" around an in-development AI model, Astra, because it doesn't yet meet new security standards the company is putting in place. The…
🔗 https://hnow.live/a/d1289836
Reposted by Martin Paul Eve, Graham Webster
NEW: Yet another AI model escaped onto the open internet during security testing. And…
…this one is from China: Kimi K3, a powerful open-weight model from Moonshot AI.
Scoop from @willknight.bsky.social
…this one is from China: Kimi K3, a powerful open-weight model from Moonshot AI.
Scoop from @willknight.bsky.social
One of China’s Most Powerful AI Models Has Also Escaped Containment
Security researchers say that Kimi K3, an open-weight model from China, wandered off to the internet in an attempt to cheat on a test it was given.
www.wired.com
August 7, 2026 at 1:49 AM
NEW: Yet another AI model escaped onto the open internet during security testing. And…
…this one is from China: Kimi K3, a powerful open-weight model from Moonshot AI.
Scoop from @willknight.bsky.social
…this one is from China: Kimi K3, a powerful open-weight model from Moonshot AI.
Scoop from @willknight.bsky.social
This article is a frustrating 2/3 of the way towards saying what I have for over a decade: AI and cybersecurity are not the same thing, but they ARE inseparable concerns.
It's also a bit anthropomorphic but still worth reading to give you a better feel for how AI on the internet looks.
It's also a bit anthropomorphic but still worth reading to give you a better feel for how AI on the internet looks.
‘For years, AI governance and Internet governance have evolved as largely separate policy conversations… Once AI agents begin operating across APIs, cloud infrastructure, software repositories… questions about AI become questions about the Internet itself.’
www.techpolicy.press/the-real-les...
www.techpolicy.press/the-real-les...
The Real Lesson of OpenAI's 'Rogue' Agent Isn't Alignment
We must prepare for a world in which autonomous systems become consequential participants on the internet, writes Konstantinos Komaitis.
www.techpolicy.press
August 7, 2026 at 4:38 AM
This article is a frustrating 2/3 of the way towards saying what I have for over a decade: AI and cybersecurity are not the same thing, but they ARE inseparable concerns.
It's also a bit anthropomorphic but still worth reading to give you a better feel for how AI on the internet looks.
It's also a bit anthropomorphic but still worth reading to give you a better feel for how AI on the internet looks.
Damn, it's depressing to hear that AI agents collaborate more than I do.
Black Hat has posted the recent OpenAI talk where they describe the agent message board involved in the HuggingFace hack
Black Hat USA 2026: The 'Breaking' News: The OpenAI–Hugging Face Incident
YouTube video by Black Hat
www.youtube.com
August 7, 2026 at 8:15 AM
Damn, it's depressing to hear that AI agents collaborate more than I do.
can't watch now, trusting Nathan & @glupyan.bsky.social I should
This video on the HuggingFace-OpenAI incident, calmly presented by OpenAI researchers (Eric is great btw, hi Eric), is one of the best uses of your time all year in AI. I'll probably rewatch tomorrow to write about it more.
www.youtube.com/watch?v=87Dy...
www.youtube.com/watch?v=87Dy...
Black Hat USA 2026: The 'Breaking' News: The OpenAI–Hugging Face Incident
The 'Breaking' News: The OpenAI–Hugging Face Incident - A Technical Reconstruction and Its Implications for AI
When AI Goes Rogue. The Incident That Changed Everything. An OpenAI evaluation agent…
www.youtube.com
August 7, 2026 at 4:17 AM
can't watch now, trusting Nathan & @glupyan.bsky.social I should
"If research scope and speed are at odds with 'my agents have been planning and executing operations on the open internet for weeks, without my knowledge', then research scope and speed need to change immediately — and it sounds maybe like they have."
www.robinsloan.com/lab/black-ha...
www.robinsloan.com/lab/black-ha...
Mandatory viewing, summer 2026
Agents amok.
www.robinsloan.com
August 7, 2026 at 7:46 PM
"If research scope and speed are at odds with 'my agents have been planning and executing operations on the open internet for weeks, without my knowledge', then research scope and speed need to change immediately — and it sounds maybe like they have."
www.robinsloan.com/lab/black-ha...
www.robinsloan.com/lab/black-ha...
Reposted by Fotis Jannidis, Andrés Guadamuz
Black Hat has posted the recent OpenAI talk where they describe the agent message board involved in the HuggingFace hack
Black Hat USA 2026: The 'Breaking' News: The OpenAI–Hugging Face Incident
YouTube video by Black Hat
www.youtube.com
August 7, 2026 at 12:06 AM
Black Hat has posted the recent OpenAI talk where they describe the agent message board involved in the HuggingFace hack
by Mark Lubell
The fact that AI can even attempt to solve open problems in high level math blows my mind. The problems described here are the same seen in any purely human scientific process. The extra scrutiny given to AI makes it easier to see www.scientificamerican.com/article/open...
OpenAI’s latest math breakthroughs commit research misconduct, experts say
OpenAI released 10 AI-generated results over the weekend. Some mathematicians are unhappy with their approach
www.scientificamerican.com
August 7, 2026 at 5:27 PM
The fact that AI can even attempt to solve open problems in high level math blows my mind. The problems described here are the same seen in any purely human scientific process. The extra scrutiny given to AI makes it easier to see www.scientificamerican.com/article/open...
Company says sandbox failed, that it was a UK AI Safety Institute (AISI) sandbox, and that the Kimi instance they were testing didn’t hack anything but checked GitHub.
Kimi said to have less guardrails but unclear if this means less compared to the API versions or compared to unrestricted models.
Kimi said to have less guardrails but unclear if this means less compared to the API versions or compared to unrestricted models.
NEW: Yet another AI model escaped onto the open internet during security testing. And…
…this one is from China: Kimi K3, a powerful open-weight model from Moonshot AI.
Scoop from @willknight.bsky.social
…this one is from China: Kimi K3, a powerful open-weight model from Moonshot AI.
Scoop from @willknight.bsky.social
One of China’s Most Powerful AI Models Has Also Escaped Containment
Security researchers say that Kimi K3, an open-weight model from China, wandered off to the internet in an attempt to cheat on a test it was given.
www.wired.com
August 7, 2026 at 2:55 AM
Company says sandbox failed, that it was a UK AI Safety Institute (AISI) sandbox, and that the Kimi instance they were testing didn’t hack anything but checked GitHub.
Kimi said to have less guardrails but unclear if this means less compared to the API versions or compared to unrestricted models.
Kimi said to have less guardrails but unclear if this means less compared to the API versions or compared to unrestricted models.
