OpenAI now says "entirely automating everything is not the future we want"

A new blog post from CEO Sam Altman and chief researcher Jakub Pachocki strikes a much more cautious tone.

Instead of full automation, the post says, "Our internal belief is that by March of 2028 we may have a significant fraction of our research being done by AI systems in tandem with our own researchers."

Whether this shift has technical or societal reasons is an open question. But Altman and Pachocki clearly position themselves against fully automating human tasks, framing it as a societal concern.

"Entirely automating everything is not the future we want. It would be unfulfilling, and it would be dangerous. […] A key long-term role for people will be deciding what is worth doing."

The more capable AI systems become, the more important the human role gets: setting direction, making tradeoffs, applying judgment.

OpenAI wants an international body that could slow AI development

At the same time, OpenAI argues that "AI doing AI research will become the determining factor of the pace of progress within the next few years." Much like Anthropic did a few days ago, OpenAI is now floating the idea of slowing AI development. Anthropic had called for the option to pause.

Altman and Pachocki propose an international organization to coordinate leading AI efforts and reduce catastrophic risks. It should enable "coordinated action, including slowing frontier development when needed, so societal resilience, safety, and alignment can keep pace."

The demand sits somewhat awkwardly alongside OpenAI's own very ambitious goals. In the same blog post, the company outlines three core goals: "Build an automated AI researcher," "Accelerate the economy," and "Give everyone on Earth a personal AGI."

Altman and Pachocki describe the current moment as the start of OpenAI's third phase. The first phase focused on foundational research. The second built the company into a product business. Now the goal is to make advanced AI "abundant, affordable, safe, useful, and easy enough" for every person and organization to benefit.

"Frontier capability is only part of the job. The bigger task is turning that capability into tools people can actually use to thrive," they write. That thinking lines up with OpenAI's broader strategic shift from model provider to implementation partner, the same logic behind its new DeployCo subsidiary, which sends engineers directly into companies to embed AI into their workflows.

It's also a tacit admission that AI implementation is far harder than just offering a chatbot and that adoption may not move fast enough, with clear enough ROI, to support the revenue growth OpenAI and Anthropic will need in the coming months.