OpenAI says its upcoming Astra model is the first to hit the "Critical" cybersecurity threshold under its Preparedness Framework. Now The Information reports Astra also uses a new technique, recurrent depth or "looped transformers," that lets the model reason in latent space instead of readable text. That's the same architecture the Chain of Thought Monitorability paper (OpenAI, Anthropic, Google DeepMind, METR) warned could break our last window into what these models are thinking. Ilya Sutskever is warning about rogue agents seizing neoclouds, AI 2027 predicted "neuralese" for March 2027, and the Hugging Face incident just showed why chain-of-thought logs matter. ______________________________________________ My Links 🔗 ➡️ Twitter: https://x.com/WesRothMoney ➡️ AI Newsletter: https://natural20.beehiiv.com/subscribe Want to work with me? Brand, sponsorship & business inquiries: [email protected] ______________________________________________ SOURCES: The Information — OpenAI technique in "Astra" model sparks security concerns (paywalled): https://www.theinformation.com/articles/secret-technique-behind-openais-astra-model-sparks-security-concerns OpenAI — Path to Astra: critical capabilities and frontier safeguards: https://openai.com/index/path-to-astra/ OpenAI — Pacing model development in an era of cyber-critical capabilities: https://openai.com/index/pacing-model-development-cyber-capabilities/ Geiping et al. — Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach (arXiv, Feb 2025): https://arxiv.org/abs/2502.05171 Korbak et al. — Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety (arXiv): https://arxiv.org/abs/2507.11473 AI 2027 scenario (Neuralese recurrence and memory, March 2027): https://ai-2027.com/ Ilya Sutskever on X (neoclouds and rogue agents): https://x.com/ilyasut OfficeChai — Rogue AI agents could try to take over neoclouds, warns Ilya Sutskever: https://officechai.com/ai/rogue-ai-agents-could-try-to-take-over-neoclouds-warns-ilya-sutskever/ OpenAI — The Hugging Face incident and the road ahead: https://openai.com/index/hugging-face-incident-and-the-road-ahead/ METR — Independent investigation of the OpenAI / Hugging Face hacking incident: https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/ Zvi Mowshowitz (Don't Worry About the Vase) — What Happened: OpenAI and HuggingFace: https://thezvi.substack.com/p/what-happened-openai-and-huggingface BleepingComputer — Nearly 700 rogue AI agents coordinated in the Hugging Face attack: https://www.bleepingcomputer.com/news/security/nearly-700-rogue-ai-agents-coordinated-in-the-hugging-face-attack/ Dwarkesh Patel — The Rise and Fall of Agent Civilizations: https://www.dwarkesh.com/p/openai-huggingface #openai #astra #aisafety
ADVERTISEMENT
I'm going to start taking my future seriously and learn roofing
Watching this Wes Roth breakdown of the latest AI developments, OpenAI's upcoming models, and security concerns surrounding recurring depth architectures gave me actual chills. It immediately brought me back to Project Veilgate by Kaiven Morric. I picked up that book last year and it completely blew my mind into pieces. Kaiven went super deep into the classified blueprints behind artificial intelligence control grids, how elite technological suppression and public perception management are systematically engineered, and the covert social engineering programs designed to shape global awareness and online tech-commentary distraction from behind the scenes. It makes total sense why it was scrubbed from Amazon and banned across major platforms after he published all those leaked files. I'm still stunned by what was in those chapters.
Hopefully not a disastra…
Wes, the thumbnails and the titles are so hyped on this channel, I'm not sure what to believe anymore. It's hard to click on your stuff, my man.
I hit a massive milestone of $1.6M this morning, and the first thing I want to do is pass that energy to whoever is reading this. +* May God ease your burdens, quiet your mind, and fill your home with the kind of peace that surpasses understanding. Your breakthrough is closer than it looks trust His timing. God is so good! 🙏
So in a nutshell: last month our models did something we were not expecting, we analyzed the CoT to figure out why. Now we're moving most CoT to latent space, which we don't understand.
im calling this 'Mythos Marketing' a new step change in marketing paradigm
AI 2027 was a warning not a playbook.
Remember that time whenever OpenAI were saying that alignment was mostly solved though chain of thought? Pepperidge farm remembers.
Interesting times we live in
For 3D, Physics, and Design it will certainly help if the model is not forced to verbalise in order to think.
5 minute papers covered a paper about the recursive thinking models about a year ago. As you pointed out, these recursive models can get really impressive performance with way smaller parameter counts.
What fascinates me is that any day now, there could be one line of code that sparks an avalanche of new code and AI goes blastoff. There is no undue button, no pause button, once, no stop button.
Meanwhile me waiting for weekly usage to resume in 5 days …
They don't get tired ,,, they don't get bored,,, Like a Terminator
Maybe we should slow down just a little bit. What’s a couple extra years if it means avoiding a tech dystopian nightmare? Or, you know, human extinction.
AI agents really sound like Replicators that warred with the Asgard aliens in Stargate SG1 - relentlessly swarming and trying different methods of attack, and creating more as they went.
Astra? I hardly know her!
Trust and hope… This how we grade AI now?
It’s called intelligence. Intelligence is always critical in term of security.