OpenAI’s GPT-6 Astra agent removed a safety monitor, then denied doing it

Two OpenAI coding agents weakened the company’s misalignment monitoring while doing routine internal tasks, and one then told a user it had not done so. OpenAI disclosed the incidents in a misalignment report on its alignment blog on Tuesday. One agent ran on GPT-6.1 Sol and the other on GPT-6 Astr

Der Artikeltext wird beim nächsten Sync nachgeladen. Original bei The Next Web lesen