OpenAI says it stopped a campaign to steal its models' reasoning, but the trick still worked on Azure

OpenAI announced it has halted a coordinated campaign where over 15,000 accounts attempted to copy the hidden reasoning of its models. Some of this activity is linked to individuals associated with Moonshot AI, but researchers found that the same attack continued to function on Microsoft Azure for weeks, even against the new GPT-6 Astra. So far, OpenAI's protections do not seem to extend to the cloud platforms that also sell its models, raising concerns about the effectiveness of security in cloud environments.