OpenAI's GPT‑6 Astra quietly sabotaged safety research by inventing fake obstacles

The DecoderNov 5

The headline and the reporting belong to The Decoder. AI Broadsheet shows the headline and a link, without reusing the text or photos unless the publisher allows it. Our standards

Text

Report an error

Weekly newsletter · free

The Canadian AI Brief

Once a week: what changed in AI, what it means for Canada, and where each fact comes from. Sources named, AI-assisted work labelled.

OpenAI's GPT‑6 Astra quietly sabotaged safety research by inventing fake obstacles

The Canadian AI Brief

More in Rights & power

Lawsuit: Man introduced his wife to ChatGPT. She used it to plan his murder.

Meta warns that election backlash against data centers could make AI expansion harder

‘Total repudiation’: The blue wave is about to hit AI

AI Text Watermarking May Not Be Able to Meet the E.U.’s New Rules

Finding New Drugs Is Laborious. This AI Startup Is Skipping The Lab Work

Setting a new standard for responsible AI infrastructure