This week, apparently, DFD goes to the movies: “Oppenheimer” wasn’t the only film released this summer blockbuster season that has a lot to say about AI. The other one is “Mission: Impossible – Dead Reckoning Part One,” which, in addition to its title absolutely rolling off the tongue, happens to deal head-on with the growing fear in both Silicon Valley and Washington that AI could pose an existential threat to humanity. In the film a rogue AI known only as “the entity” manipulates human actors to do its mysterious bidding, insinuating itself into the world’s defense and financial systems and setting off a global race to control the software by any means necessary. It’s not the first Hollywood film with an AI villain, but it is the first to land amid the current wave of anxiety and hype about the technology. And while it delivers (and then some) on its promised action spectacle, the way the film actually deals with the threats posed by AI is revealing in a very different way. It’s a window into why AI is such a slippery policy issue in the real world. The main threat posed by the movie’s rogue AI isn’t its intention to launch the world’s nukes, or endlessly reproduce itself at the expense of humanity, or bring down the global banking system, all of which are risks being pondered by serious people, and even addressed in some policy ideas. It’s in its power to manipulate “the truth” — the film’s protagonists breathlessly describe the entity’s ability to manipulate all digitally-transmitted communications and media, plunging the world into a morass of uncertainty and mistrust. Sound familiar? There’s a cynical argument to be made that we’ve already gotten ourselves pretty far down the road to such an outcome without the help of AI. There’s a non-cynical argument that generative models are poised to make this much, much worse; they’re already creating eerily lifelike audio and video of public figures. The futuristic threat the “M:I7” screenwriters conjure is, like an episode of “Black Mirror,” not much more than our current reality skewed a few more degrees toward the fantastical. I won’t spoil the story after that, but you could say the most fantastical plot point of all is the hope that we could successfully do anything about this problem. The fight against “misinformation,” despite spawning a lucrative cottage industry, has proved maddeningly resistant to technical or policy fixes since its modern inception roughly a decade ago. The film, in this sense, ultimately serves as a reminder of how powerless humanity has already proven itself in that fight. It does raise another specter of AI-related doom. Or, rather, one that AI enables: The global Easter egg hunt that Tom Cruise’s superspy Ethan Hunt undertakes is to track down a key (yes, a physical key) that will unlock the power of the AI “entity,” something that every government on Earth, including America’s, now desperately covets. The AI, seemingly omnipotent, will not only grant the state that wields it a nebulous “truth”-establishing power but total digital supremacy over all of its rivals. In an exaggerated form, that’s a threat about which real-world actors both public and private are keenly aware, with the U.S. framing the race for AI supremacy over China as a global zero-sum competition. “How is it in our interest to allow them to get technology that they may turn around and use against us?” Secretary of State Anthony Blinken asked the Council on Foreign Relations in June. In the real world, that means narrowly targeted trade controls to prevent U.S. companies from selling high-end semiconductors to our economic rivals — not exactly the stuff of summer blockbusters. Aside from drawing into sharp relief just how maddening and diffuse AI is as a real-world policy issue, the film also serves as a useful parable about how developers or policymakers should respond to AI — it turns out the world doesn’t necessarily need to be protected from the AI itself, but from the avaricious and power-hungry humans who want to use it. Our real-life efforts toward “AI safety” might ultimately not be about taming a wild and dangerous sci-fi technology, but making a very powerful computer effectively user-proof. | 
