ResearchSep 2, 2026 · 09:578 sources

OpenAI Adopts Loop Transformers in Astra to Block Chain of Thought Monitoring

From huggingnews.com

Image: huggingnews.com
AI;DR

OpenAI's upcoming Astra AI model uses a recurrent depth architecture, also known as loop transformers, to improve processing costs and performance. This method obscures the internal reasoning process, making it more difficult for safety researchers to monitor the chain of thought. OpenAI is reportedly limiting the use of this technique to maintain oversight as the model reaches a critical cybersecurity threshold.

Read the sourceSign in to suggest

Sources

+ 6 more
  1. SUPPORT@amir“a leap forward on performance, but sparking concerns inside & outside OpenAI re: security”x.com
  2. SUPPORT@_nathancalvin“breakthrough in neuralese for Astra that could destroy chain of thought monitorability”x.com
  3. SUPPORT@_nathancalvin“monitorable chain of thought as 'A New and Fragile Opportunity for AI safety'”x.com
  4. SUPPORT@wallstengine“found and chained together two zero-day vulnerabilities”x.com
  5. SUPPORT@zeffmax“confident it can release Astra safely”x.com
  6. SUPPORT@tradfi“limit Astra’s full cyber abilities to test group”x.com

One ranked edition, every morning

The stories that matter today. Free, cancel anytime.

Telegram @aidr_today