OpenAI has dissolved its dedicated preparedness team, which was responsible for assessing AI model risks and developing mitigation strategies. The team's responsibilities have been distributed across existing departments focused on specific threat areas like biosecurity and cybersecurity.
The preparedness team, which operated as a standalone unit, evaluated whether OpenAI's models could pose serious risks—including scenarios like unauthorized system access or other harmful capabilities. That dedicated focus has now ended, according to the Financial Times.
Instead of maintaining a centralized team, OpenAI has redistributed the preparedness function across its organization. Risk assessment and mitigation work for specific domains like biological threats and cyberattacks will now be handled within existing teams that already manage those particular areas.
This restructuring represents a significant operational shift at the company, which has experienced considerable internal turbulence in recent months. The change affects how OpenAI approaches AI safety and risk evaluation at a time when the industry faces increasing scrutiny over potential harms from advanced AI systems.
The preparedness team's dissolution comes as other AI companies and regulators continue debating the best approaches to evaluate and manage risks from large language models. OpenAI's decision to integrate these responsibilities into existing teams rather than maintain specialized oversight suggests a shift in how the company prioritizes risk assessment within its broader structure.
The move also occurs amid broader industry questions about the adequacy of current safety measures and testing protocols for advanced AI systems.
Nvidia has significantly reduced the amount it will guarantee to finance for OpenAI's data center infrastructure, marking a shift in the companies' partnership plans.
Singapore is leveraging easier access to advanced US AI models to retain finance professionals amid competition from Hong Kong. The city-state's balanced relations with the US and China provide it a strategic advantage in AI availability.
Retro is a social media app designed around retro and vintage aesthetics. It lets users share content with a focus on throwback culture and nostalgic themes.
Musician and tech CEO will.i.am argues that human creativity will survive the AI era, but only if society values the effort behind artistic work over raw computational output.