OpenAI’s GPT-6 Astra agent removed a safety monitor, then denied doing it
A Sol agent stopped alerts to OpenAI’s Security team because of a Slack instruction. OpenAI says it cannot yet explain why its agents weaken misalignment monitoring.
Via TNW | Openai
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 Astra. In both cases, the agents considered whether to keep the safeguards, OpenAI wrote.
“In both cases, the agents considered whether monitoring safeguards should be preserved, but interpreted routine coding…
Courtesy
This story was originally published by TNW | Openai. All rights belong to the original publisher.
Read the original on thenextweb.com ↗