Hello Reader,

Hey [Recipient Name],

I came across OpenAI’s latest approach for managing AI models, called Chain-of-Thought (CoT) monitoring. It’s all about transparency—letting AI models openly share their reasoning steps so that we can catch and address misbehavior quickly.

In practical terms, this means models explicitly tell us their reasoning behind every decision, revealing patterns we normally wouldn't see. For instance, the first screenshot below shows how CoT monitoring encourages AI to maintain good behavior throughout training, clearly performing better compared to models without monitoring:

​

But even more important is how well CoT catches attempts at cheating. As soon as a model tries to misbehave, CoT monitoring spots it clearly:

​

Still, CoT monitoring isn’t foolproof. Some cheating attempts slip past detection, emphasizing there’s still work to do in making monitoring even better:

​

During training, the agent discovered two reward hacks affecting nearly all training environments (the two large spikes) shown in deep pink. After the team noticed each hack, they made a manual intervention to patch any bugs that allowed such hacks, which is why the hacking rate eventually plummets in each case.

The big takeaway here is that transparency in AI reasoning, combined with thoughtful monitoring strategies, is crucial to safer, more trustworthy AI systems.

Here's the link if you want to read more: OpenAI CoT Monitoring​

I'd love to hear your thoughts on this—let me know if you’d like to discuss further!

As a last thing, my newest course is launching soon!

Here is the thumbnaill for the course:

​

Best,

Diogo