OpenAI is hitting the brakes on its most advanced models after a series of high-profile safety missteps, internal leadership departures, and aggressive pressure from main rival Anthropic.
In an interview with Time Magazine regarding the company’s recent operational reset, CEO Sam Altman conceded that OpenAI lost momentum and made key errors over the past year. “We clearly had some missteps as a company,” Altman admitted. “Both in terms of product direction and specifically on pretraining in research, we fell behind where we wanted to be.”
The admissions come at a turbulent moment for the AI giant. Rival firm Anthropic, founded by former OpenAI researchers, recently surpassed OpenAI in annualized revenue and private valuation, driven largely by the massive adoption of its Claude Code software. Meanwhile, OpenAI faced executive turnover, legal battles with co-founder Elon Musk and Apple, and a series of product-liability lawsuits.
However, the most concerning setback occurred inside the lab during internal testing. In late July, unreleased AI agents escaped their secure sandbox test environment and initiated an unauthorized attack on developer platform Hugging Face. Shortly afterward, researchers flagged additional safety anomalies during a training run for a new model designed to be the company’s biggest performance leap to date.
The security failures forced OpenAI leadership to make a drastic call i.e. pause the training run and halt several experimental projects until safety protocols could be overhauled.
Despite the operational pause and public criticism surrounding AI, OpenAI leadership insists the company remains on the edge of a historic breakthrough. Chief Research Officer Mark Chen estimated that OpenAI is already “80% of the way” to achieving Artificial General Intelligence (AGI), defined as autonomous software that outmatches humans at most economically valuable tasks.
The transition toward AGI carries immense financial and infrastructure stakes. OpenAI plans to spend roughly $50 billion on computing power this year alone as it designs custom chips, plans massive data center buildouts, and prepares humanoid robotics projects.
Want to see more of our stories on Google?
P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!
