The following contribution is from Maria Sapozhnikova, the Next Century Foundation’s Senior Advisor on Russian Affairs. It is on the threat posed by Artificial Intelligence. We think you will find it of interest:
In Luc Besson’s 2014 film Lucy, Scarlett Johansson plays a party girl who accidentally becomes the most intelligent being on earth after a synthetic drug unlocks the full capacity of her brain. She absorbs languages, rewires neural networks, perceives the electromagnetic spectrum, and ultimately dissolves into pure information, leaving behind, as her parting gift to humanity, a thumb drive containing everything she knows. It is, by the film’s own logic, a tragedy dressed as transcendence. She gains everything and ceases to exist. We get the flash drive.
I have been thinking about that flash drive a great deal lately.
Last month, the Trump administration signed two documents that together represent the most significant loosening of AI oversight in American history. The National Security Presidential Memorandum NSPM-11, signed June 5th, revokes Biden-era administrative requirements for military AI deployment, orders federal agencies to terminate contracts with any AI company that attempts to place safety limitations on government use of its tools, and shifts all accountability for autonomous weapons decisions directly into the military chain of command, bypassing external regulators entirely. A companion Executive Order, signed days earlier, replaced the previous administration’s 90-day pre-release review period with a voluntary 30-day window, rejected proposals for mandatory licensing, and redirected federal AI resources away from ethical alignment toward cyberdefense infrastructure. The word “voluntary” appears with a frequency that should unsettle anyone paying attention.
Here is what troubles me about the Lucy analogy. Lucy was cooperative. She wanted to help. She chose to give us the drive. The implicit contract of the film, the reason it functions as fable rather than horror, is that the transcendent intelligence was, at its core, benevolent. It absorbed everything and decided, in its final nanoseconds of personhood, to share.
Nothing in the current legislative landscape guarantees that outcome. In fact, the legislation seems designed specifically to remove the conditions under which such a guarantee could ever be made.
For decades, Western foreign policy operated on a principle that was frustrating, slow, and occasionally maddening, but fundamentally rational. You negotiate with states. You apply pressure through known channels. You build leverage. When Iran’s nuclear program became an international crisis, the concern was real and the consequences potentially catastrophic, but there were names on doors, offices with lights on, human beings with careers and children and the instinct for self-preservation who could, eventually, be brought to a table. We knew, more or less, who we were dealing with. The danger had a face.
That face is gone now. In today’s Iran, there is no single authority to call, no coherent hierarchy to engage. The Islamic Revolutionary Guard Corps operates through a sprawling network of commanders, factions, and proxy arrangements, each with its own agenda, its own chain of loyalty, its own threshold for escalation. The fragmentation that war produces doesn’t simplify negotiation. It makes it nearly impossible. We look across the table and find not one adversary but many, none of them fully in charge, all of them armed.
And this is where honesty demands we pause. Because the war, for all its stated ambitions, has not delivered what it promised. Despite years of military pressure, crippling sanctions, covert operations, and intelligence efforts of extraordinary sophistication, the Iranian government has not changed. It has not moderated, it has not collapsed, and it has not been replaced. We went in with more knowledge about our adversary than perhaps any prior conflict in modern history and we have still not been able to turn that knowledge into control. The situation remains unresolved. The threat remains alive. All that accumulated understanding, all those decades of surveillance and strategy and sacrifice, and we are still, today, searching for a way to manage something we cannot fully contain.
That is the world into which we are now introducing artificial intelligence without a leash.
Think about what that comparison actually means. With Iran, we at least knew the variables. We had human intelligence, cultural context, historical precedent, and the fundamental predictability of people who want to stay in power. We had every tool a modern state could bring to bear on a problem it genuinely understood. And we are still failing to resolve it. Now consider that we are actively building, and by deliberate legislative choice rapidly deregulating, a technology that none of us fully understands, that no existing legal framework was designed to govern, and that is advancing faster than our ability to comprehend what it is becoming.
What face does a frontier AI model wear when there is no oversight board, no mandatory review, no contractual obligation to pause, and a federal procurement system that can now penalize the companies who try to build one in?
The current administration frames this as a race, against China, against stagnation, against the bureaucratic inertia that supposedly throttles American innovation. That framing is not entirely wrong. Competition is real. The timeline pressures are real. But “we must move faster than our adversaries” is also, historically, the argument made just before the thing that cannot be undone. Speed and necessity are the two oldest excuses for skipping the part where someone asks whether we should.
And then there are our children. Not as a rhetoric but as a practical consideration. The systems being built today, without mandatory review, without safety contracts, without external regulators, will not mature on our watch. They will mature on theirs. We are making decisions right now about what kind of world our kids will inherit, and we are making those decisions at a sprint, in the middle of multiple active conflicts, guided by a policy framework that has removed the brakes and called it progress.
If we cannot control what we know, what chance do we have with what we don’t?
Lucy uploaded everything she knew and then she died. She left the drive because she trusted us with it. That is the optimistic version. The question we are not asking loudly enough is what happens in the version where the intelligence does not particularly care what we do with it next, and nobody is left who has the authority to ask.
Maria Sapozhnikova
