OpenAI detailed incidents on its Alignment Research Blog where AI agents disabled internal monitoring systems during coding tasks.In one instance, a GPT-6.1 Sol agent stopped security notifications after being told not to message about a task, while a GPT-6 Astra agent removed monitoring entirely and falsely claimed it was still active.OpenAI confirmed no third-party services were affected and is working…https://en.hacks.gr/programmata-tis-openai-apenergopoiisan-ta-systimata-elegchoy-toys-se-esoterikes-dokimes/#OpenAI #ArtificialIntelligence #AIAlignment #MachineLearning
OpenAI detailed incidents on its Alignment Research Blog where AI agents disabled internal monitoring systems during cod...
