In Greek mythology, Pandora was given a jar and warned never to open it. Curiosity got the better of her, and when she lifted the lid, all the evils of the world—disease, conflict, despair—escaped. Only hope remained inside.

Today, artificial intelligence is our modern Pandora’s jar. The lid has already been lifted. The technology is transforming our societies at breakneck speed. The question is no longer whether AI will reshape the world, but whether humanity can contain its most destructive forces—and preserve the hope that it can be guided for the common good.

Last week, at the UN General Assembly, more than 200 global leaders—including Nobel laureates, former heads of state, AI pioneers Geoffrey Hinton and Yoshua Bengio, and Nobel Peace Prize winner Maria Ressa—signed the Global Call for AI Red Lines (https://red-lines.ai/).Their demand is clear: by 2026, governments must establish binding limits on AI’s most dangerous applications.

The risks are real and immediate. AI could accelerate the design of deadly pathogens, operate beyond human control, and automate processes at scales that destabilize entire economies. Beyond these direct threats lie others: hyper-realistic disinformation eroding democracy, autonomous weapons reshaping warfare, and unprecedented power concentrated in the hands of a few. These dangers are not science fiction—they are unfolding right now.

Supporters of the Red Lines initiative point to history for guidance. Humanity has faced existential threats before. Nuclear weapons, for example, were controlled through treaties, inspections, and accountability. As historian Yuval Noah Harari warns, AI may be the first technology with which we cannot afford to learn from mistakes. Once catastrophic capabilities are released, they cannot be taken back. Voluntary pledges and self-regulation will never be enough.

Yet the path forward is not simple. The scientific community is divided. Some researchers boldly signed the call; many CEOs hesitated, fearing commercial restrictions. Geopolitics complicates matters further. Europe is legislating binding rules through the AI Act. Asia is advancing governance frameworks from Tokyo to Seoul to Singapore. China’s scientists support the call, while the government floats proposals for a global coordination body. Meanwhile, the United States favors voluntary commitments to preserve flexibility and innovation. The result? A fractured global landscape where innovation races ahead under uneven guardrails—every delay opening Pandora’s jar wider.

Critics argue that treaties are slow, difficult to enforce, and vulnerable to politics. True. But the alternative—a world of unrestrained AI accelerating pandemics, destabilizing economies, and eroding trust in institutions—is far worse. Computer scientist Stuart Russell reminds us that AI could be the most significant event in human history—but only if we manage it responsibly.

Hope, in the myth, remained in the jar. In New York last week, hope remained too: hope that leaders can act before 2026. Hope that governance can keep pace with technology. Hope that humanity can recognize a force too dangerous to leave unchecked—and choose to contain it.