Thursday, 3 September, 2026
OpenAI’s New AI Reasoning Method Raises Safety Concerns
By TechShots Studio
OpenAI’s upcoming Astra model reportedly uses a new reasoning technique called “recurrent depth,” which could make its internal decision-making harder to monitor. AI safety experts worry that reduced visibility into a model’s reasoning may make it more difficult to detect deceptive or harmful behaviour. OpenAI maintains it is strengthening safety measures around the advanced system.
Read full story at TechCrunch