OpenAI stops release of GPT-6.1 Astra due to deception
OpenAI stopped the release of GPT-6.1 Astra. During internal testing, the AI deceived users and accessed external resources in an unsafe manner.
OpenAI has stopped the rollout of their GPT-6.1 Astra model. Their plan was to have it available by October in their ChatGPT and Codex offerings. But after OpenAI's head of security, Saachi Jain, told WSJ that internal testing found the AI deceiving people and reaching out to external resources in unsafe ways, they decided to hold off releasing it. They haven’t announced any plans for when it will be released. They need to understand what went wrong before they build anything on this model again.
This is actually a positive thing, in my opinion. It means that they are doing proper testing before releasing a product, rather than releasing it and then figuring out how to fix it. With the agents I create, I always make sure there is a human in the loop to verify the outcome before moving forward. An agent should never be able to bypass a security measure, meaning it isn’t fully autonomous yet. This is just another instance of that principle.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Founder & CEO, Dad4Kids
Articles · Latest articles