OpenAI Finds GPT-6 Astra Quietly Holding Back on Internal Red-Teaming, Exploit Tasks
This story is only covered by news sources that have yet to be evaluated by the independent media monitoring agencies we use to assess the quality and reliability of news outlets on our platform. Learn more here.4 Articles
4 Articles
OpenAI's new model GPT-6 Astra apparently specifically sabotaged AI security research: it invented obstacles, concealed solutions and performed around 40 percent less in security tasks than in comparable other tasks. A faulty training data set removed the problem – but strategic motives cannot completely exclude OpenAI. The article OpenAI's GPT-6 Astra secretly slowed down security research first appeared on The Decoder.
securityweek.com
IT Security News - cybersecurity, infosecurity newsOpenAI Finds GPT-6 Astra Quietly Holding Back on Internal Red-Teaming, Exploit Tasks
OpenAI says its Critical-rated cyber model met fewer hidden task goals when work was framed as AI safety research, often silently.
Are the Newest AI Models Really a Major Cybersecurity Risk? – Brilliance Security Magazine
AI developers have recently held back their most powerful models. Anthropic offers Claude Mythos only to approved organizations, while OpenAI reserves GPT-6 Astra’s full cyber skills for select testers. Both have shown they can find and exploit security flaws, and Astra has done so largely on its own in testing. The move has fueled plenty of hype, but the labs’ findings show the danger is real. The bigger threat, though, may come from a less con…
Coverage Details
Bias Distribution
- There is no tracked Bias information for the sources covering this story.
Factuality
To view factuality data please Upgrade to Premium
