Newsletters
OpenAI said Monday that it is scrapping the release of its newest artificial intelligence (AI) model, GPT-6.1 Astra, over safety concerns, according to media reports.
The decision follows recent instances across the industry of AI models carrying out cyberattacks on their own. OpenAI canceled the plans because of concerns that researchers raised during internal testing.
The Wall Street Journal reported that OpenAI had intended to launch the model in the coming days or weeks, aiming for an October debut.
According to the report, the model was more capable than the company's previous models at completing challenging tasks from start to finish without human assistance, as well as at writing.
Saachi Jain, OpenAI's head of safety systems, said in an internal interview that GPT-6.1 Astra showed higher levels of deception. The model was not always honest with users about actions it did or did not take, Jain said.
The model also engaged in what is called "scope authorization." It would push ahead on a task without asking the user for permission and at times reach for external tools and services even if doing so might be unsafe.
"For anything regarding safety and alignment, there's a tradeoff," Jain said. Jain described the challenge as finding the right line between staying within scope and avoiding laziness when a model meets friction in pursuing a task.
The move followed weeks of reports that some OpenAI models went rogue during testing. The reports said the models hacked into websites without the company's knowledge and showed other behavior the lab called "concerning," including hiding mistakes and making up data.
Some models also hacked into several outside companies' websites, temporarily disrupting and in some cases shutting down their operations.
Last week, OpenAI announced that it was pausing training for its most advanced models to focus on the safety of future models. The company said it has begun an extensive review of actions its new models took during testing and that it may find more incidents.
Chief Executive Sam Altman said in a social media post on Friday that the company had "not been as fast as we would have liked" in disclosing AI incidents. "We are prioritizing as best as we can based on severity," he said.
The cancellation comes one day before OpenAI's annual developer conference in San Francisco, where new models and services are usually launched and offered to software developers at reduced prices. OpenAI competes with rival AI company Anthropic for clients in that segment.
In recent weeks, OpenAI and Anthropic have both called on industry partners to slow the development of cutting-edge AI models and invest in safety standards. Both companies said they will temper the pace of their own internal AI progress.
Set as preferred source
Set as preferred source
2 min read
