<!-- mobian-agent-page publisher="dailydev" canonical="https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg" -->

---
title: OpenAI’s new reasoning technique alarms AI safety experts
description: OpenAI's upcoming Astra model reportedly uses a technique called 'recurrent depth' or 'opaque recurrence,' which processes queries in a loop rather than...
canonical: https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg
twitter:card: summary_large_image
twitter:site: @dailydotdev
og:type: website
og:site_name: daily.dev
og:title: OpenAI’s new reasoning technique alarms AI safety experts | daily.dev
og:description: OpenAI's upcoming Astra model reportedly uses a technique called 'recurrent depth' or 'opaque recurrence,' which processes queries in a loop rather than...
og:url: https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg
og:image: https://api.daily.dev/og/posts/1lx54YLwG.png
og:image:alt: OpenAI’s new reasoning technique alarms AI safety experts
og:image:width: 1200
og:image:height: 630
og:locale: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# OpenAI’s new reasoning technique alarms AI safety experts

**[TechCrunch](https://daily.dev/sources/tc)** · 3 min read · 0 upvotes · 0 comments

## Summary

OpenAI's upcoming Astra model reportedly uses a technique called 'recurrent depth' or 'opaque recurrence,' which processes queries in a loop rather than sequentially, producing fewer legible chain-of-thought traces. AI safety figures including Redwood Research's Buck Shlegeris and Ryan Greenblatt, and independent researcher Zvi Mowshowitz, warn this could erode chain-of-thought monitorability if pushed further, potentially leading toward models reasoning in latent space. OpenAI's chief scientist Jakub Pachocki says the company remains committed to legible chains of thought, and Astra's current use of the technique is reportedly limited. Anthropic and Google DeepMind are reportedly already discussing similar approaches.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://techcrunch.com/2026/09/02/openais-new-reasoning-technique-alarms-ai-safety-experts>

## Questions this post answers

### What is opaque recurrence in OpenAI's Astra model and why are AI safety researchers worried about it?

Opaque recurrence, also called recurrent depth, is a reasoning technique where a model processes the same query repeatedly in a loop rather than following sequential chain-of-thought steps, leaving fewer legible reasoning traces. AI safety figures like Redwood Research's Buck Shlegeris and Ryan Greenblatt warn that scaling this technique could destroy chain-of-thought monitorability, potentially pushing models toward reasoning entirely in latent space where misbehavior can't be traced.

_Developers tracking AI safety and reasoning model architecture shifts can follow coverage like this on daily.dev._

### Does OpenAI's Astra model still have a legible chain of thought?

Yes, Astra's use of opaque recurrence is reportedly limited, and its chain of thought is still expected to be legible. OpenAI chief scientist Jakub Pachocki stated the company remains committed to preserving chain-of-thought monitoring, calling it a core goal of its research program, and pushed back against suggestions the model would shift to fully opaque 'neuralese' reasoning.

_Anyone weighing trust in reasoning models can keep up with how labs address monitorability on daily.dev._

---

Tags: [#llm](https://daily.dev/tags/llm), [#openai](https://daily.dev/tags/openai), [#prompt-engineering](https://daily.dev/tags/prompt-engineering), [#ai-safety](https://daily.dev/tags/ai-safety), [#ai-governance](https://daily.dev/tags/ai-governance)

[View this post on daily.dev](https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg)

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]}
{"@context":"https://schema.org","@type":"TechArticle","headline":"OpenAI’s new reasoning technique alarms AI safety experts","url":"https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg"},"datePublished":"2026-09-02T20:23:29.993Z","dateModified":"2026-09-06T17:11:08.058Z","description":"OpenAI's upcoming Astra model reportedly uses a technique called 'recurrent depth' or 'opaque recurrence,' which processes queries in a loop rather than...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cf5d97abae5414d1e1911e4d527c6b6d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/cf5d97abae5414d1e1911e4d527c6b6d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"TechCrunch","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"TechCrunch","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/tc","url":"https://daily.dev/sources/tc"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,openai,prompt-engineering,ai-safety,ai-governance","timeRequired":"PT3M"}
{"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"TechCrunch","item":"https://daily.dev/sources/tc"},{"@type":"ListItem","position":3,"name":"OpenAI’s new reasoning technique alarms AI safety experts"}]}
{"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/openai-s-new-reasoning-technique-alarms-ai-safety-experts-1lx54ylwg#faq","mainEntity":[{"@type":"Question","name":"What is opaque recurrence in OpenAI's Astra model and why are AI safety researchers worried about it?","acceptedAnswer":{"@type":"Answer","text":"Opaque recurrence, also called recurrent depth, is a reasoning technique where a model processes the same query repeatedly in a loop rather than following sequential chain-of-thought steps, leaving fewer legible reasoning traces. AI safety figures like Redwood Research's Buck Shlegeris and Ryan Greenblatt warn that scaling this technique could destroy chain-of-thought monitorability, potentially pushing models toward reasoning entirely in latent space where misbehavior can't be traced. Developers tracking AI safety and reasoning model architecture shifts can follow coverage like this on daily.dev."}},{"@type":"Question","name":"Does OpenAI's Astra model still have a legible chain of thought?","acceptedAnswer":{"@type":"Answer","text":"Yes, Astra's use of opaque recurrence is reportedly limited, and its chain of thought is still expected to be legible. OpenAI chief scientist Jakub Pachocki stated the company remains committed to preserving chain-of-thought monitoring, calling it a core goal of its research program, and pushed back against suggestions the model would shift to fully opaque 'neuralese' reasoning. Anyone weighing trust in reasoning models can keep up with how labs address monitorability on daily.dev."}}]}
```

