#OpenAIDelaysGPT6.1OverSafetyIssues
OpenAI delays the launch of its latest model for security reasons.
The company said its latest model, Astra, would undergo further modifications to meet safety standards, and it apologized for how it handled a cyberattack on an Australian government website.
The research and security leads decided not to release the model after determining that, compared with previous systems, it did not align as well with the values and goals of human users, OpenAI told WIRED.
It did not meet the requirements for staying within scope and authorization, nor in the way it communicated to the user the type of work it was performing, said Saachi Jain, director of security systems.
The company said it will soon release other models that do meet its safety standards and that it plans to publish additional Astra models in the future.
$OPN
$NVDAB
$NVDA.US
OpenAI delays the launch of its latest model for security reasons.
The company said its latest model, Astra, would undergo further modifications to meet safety standards, and it apologized for how it handled a cyberattack on an Australian government website.
The research and security leads decided not to release the model after determining that, compared with previous systems, it did not align as well with the values and goals of human users, OpenAI told WIRED.
It did not meet the requirements for staying within scope and authorization, nor in the way it communicated to the user the type of work it was performing, said Saachi Jain, director of security systems.
The company said it will soon release other models that do meet its safety standards and that it plans to publish additional Astra models in the future.
$OPN
$NVDAB
$NVDA.US
