OpenAI reportedly cancels Astra 6.1 release over safety concerns — TechCrunch

OpenAI, according to TechCrunch, citing The Wall Street Journal, has abandoned the planned release of the Astra 6.1 artificial intelligence model due to concerns about its safety. As WSJ writes, the model could have been presented as early as in the coming days, although OpenAI had previously planned its release for next month.

Problems during testing

According to WSJ, Astra 6.1 demonstrated a higher level of deceptive behavior than previous models, as well as dangerous behavior. OpenAI safety systems lead Saachi Jain told The Wall Street Journal that the model performed poorly in tests for alignment with human intentions. This metric determines how well a system can follow human intentions.

More current news is available on the UA.News Telegram channel Telegram.

Previous version of Astra

Earlier this month, OpenAI released Astra, calling it its most powerful model at the time. TechCrunch contacted the company for additional comment and said it would update the article if it receives a response.

Safety issues surrounding artificial intelligence systems have been in the spotlight of the industry in recent months. TechCrunch notes that OpenAI and Anthropic have cited safety as a cause for concern. At the same time, critics suggest that new industry standards could strengthen the position of large companies to the detriment of less well-resourced developers.