So, alignment does need to be taking seriously, you're right.

But keep in mind this is a report from OpenAI about OpenAI, who have a financial incentive to present this in a certain light. Take these things with a grain of salt.

This does not mean that models are now self-aware.