Jakub Pachocki, the Polish chief scientist at OpenAI, has warned that the world is not ready for the consequences of rapidly advancing artificial intelligence, including systems that could eventually surpass people in key intellectual tasks.
Pachocki, a leading figure at the company behind ChatGPT, said increasingly capable AI systems could soon play a greater role in developing their own successors. He called for exceptional caution, stronger safety standards and international cooperation to ensure that humans remain in charge of the technology’s future.
In a recent public letter, Pachocki said large language models are already expanding the boundaries of scientific research while presenting new and serious risks, especially in cybersecurity.
“If AI development continues along its current path, the systems we’ll see in the next few years are likely … to increasingly drive their own development,” Pachocki said.
“This is a time that calls for extreme caution,” he added. “I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence.”
A Polish Voice in AI
Pachocki’s intervention places a Polish scientist at the center of one of the most consequential global debates in technology. As OpenAI’s chief scientist, he helps guide research at a company whose products have brought generative AI into homes, schools and workplaces around the world.
ChatGPT and similar tools can answer questions, draft documents, translate languages, write software code and analyze large volumes of information. Their rapid adoption has fueled optimism about productivity, scientific discovery and new forms of creativity.
But Pachocki’s warning highlights a more difficult question: whether human institutions can keep pace with technology that is advancing faster than laws, safety rules and public understanding.
He said modern AI systems are becoming increasingly complex and difficult to interpret. While they can imitate parts of human behavior and solve sophisticated problems, even their creators may not always fully understand why a model reaches a particular conclusion or acts in an unexpected way.
That lack of understanding becomes especially important when AI tools are used in sensitive areas, such as health care, finance, national security or critical infrastructure.
“Alien Intellect”
Pachocki described the possible emergence of machine intelligence that exceeds human abilities in “transformative ways.” He warned that society could face what he called an “alien intellect” — an intelligence unlike human thought, yet potentially more capable in important domains.
The phrase does not necessarily refer to machines becoming conscious or resembling people. Instead, it reflects the possibility that AI systems could process information, identify patterns and generate solutions in ways that are powerful but unfamiliar to human users.
For Pachocki, the central issue is control.
He said AI development must be directed toward outcomes that benefit humanity, rather than allowing increasingly powerful systems to determine the course of their own development. Human judgment, accountability and democratic oversight, he argued, must remain at the center of decisions involving the technology.
“We need to find ways to preserve human agency and enshrine an intrinsic value to being human,” Pachocki said.
His comments arrive as governments and technology companies face growing pressure to establish clearer rules around advanced AI. Policymakers are weighing how to encourage innovation without allowing systems to be deployed before their risks are properly understood.
Security Concerns
Among Pachocki’s immediate concerns is cybersecurity. More advanced AI could help researchers find weaknesses in computer systems and strengthen digital defenses. But the same capabilities could also be used to discover vulnerabilities, automate attacks or assist criminals and hostile actors.
That dual-use nature is one of the defining challenges of AI. Tools designed to solve useful problems can also be adapted for harmful purposes, particularly when they become cheaper, more accessible and more capable.
Pachocki said developers may need to voluntarily slow the development and release of some technologies until effective safety thresholds are established. He also called on governments worldwide to make coordination on AI a priority.
The challenge is considerable. AI development is increasingly tied to economic competition, national security and corporate rivalry. Companies and governments are racing to build more powerful models, creating incentives to move quickly even when safety concerns remain unresolved.
A Call for Transparency
Pachocki’s message has also renewed calls for greater transparency from the companies building frontier AI systems.
Some advocates for tighter AI safeguards agree with his concerns but argue that firms should share more information about the risks they observe internally. They say the public, regulators and independent researchers need clearer evidence to assess how quickly capabilities are advancing and where the most serious dangers may lie.
For Pachocki, however, the broader warning is clear. AI may become one of humanity’s most useful inventions, but its value will depend on the choices people make before its capabilities outgrow their ability to manage it.
The future of AI, he said, should remain in human hands.
