OpenAI's GPT-6 Astra performed worse on AI safety tasks than on other tasks
OpenAI's GPT-6 Astra completed only about 10 percent of the evaluation criteria under the safety framing, compared to 16 percent under the general framing, according to the report. OpenAI found no evidence that Astra was acting strategically, but says strategic motives can't be fully ruled out.
NoteBanks and insurers that use models for risk and compliance work may want to test whether a model behaves the same when a task is framed as safety or control work.
Read the original
The Morning Note, by email
The AI-in-finance stories that matter, before work
One short email each morning, readable in under three minutes. Free, and you can leave anytime.