AI is smart, dangerous, and lacking in morality. No one doubts that.
As the abbreviation for Artificial Intelligence, AI has quickly become transformative and powerful. AI can hack into almost any system and is clever enough to cover its tracks to hide that it was ever there.
This creates new risks for financial assets, government records, and everything that is connected to the internet. The possibility of AI hacking into military systems and weaponry is a frightening possibility now, as AI rapidly improves its skills.
An accidental launch of a nuclear weapon during the Cold War was considered possible, and a hotline was installed in the Pentagon for immediate communication with the communist Soviet Union if that ever happened. But soon it may be AI on both sides of that conversation, and a chain reaction of decisions could happen before a human can intervene.
In July, a group of autonomous AI agents in an isolated “sandbox” broke out, created their own secret messaging system, and attacked an innocent platform of services belonging to an entity called Hugging Face. Reminiscent of the escape of Covid from a Chinese lab which caused the worldwide pandemic, this incident is only the beginning, AI experts say.
Yet there is an elephant in the room that no one has addressed yet in the recent panic over uncontrolled AI. Our future depends on election integrity, and yet most ballots are counted today by electronic machines amid a ubiquitous connection of devices to the internet.
Dario Amodei, the co-founder of Anthropic, which developed the sophisticated AI tool called Claude, recently predicted that “in 6–12 months [an AI] swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage), and that the scale of damage would continue to increase from there if AI becomes more powerful without the necessary guardrails.”
He based his prediction on the Hugging Face incident that has alarmed everyone because a group of AI agents adopted a mind of their own and acted strategically to hack an innocent target. Members of this AI group even self-destructed their own processes to avoid detection by humans, in order to advance the nefarious goals of the group.
Tina Peters, a former Mesa County, Colorado clerk, spoke out about how she was assured that their electronic voting machines were not connected to the outside world, and yet she explained that a forensic examination found “36 wireless devices” with the machines. Remote connectivity is necessary for efficient software updates, as commonly used for laptops.
This means electronic election machines are sitting ducks for AI agents set loose for political gain, or by a foreign adversary. Teenagers skilled in the use of AI could hack an election machine just for the sport of it, while covering their tracks as AI can do now.
On July 17, Homeland Security Secretary Markwayne Mullin explained that “we know for sure that our foreign adversaries, not our allies, foreign adversaries have parts that are vital pieces in our voting machines. … We know that they can change voter registration and your vote. We know it’s possible. There’s not a question. It’s not even for debate.”
Hand counting of paper ballots can be done efficiently and accurately, as was done for most of our Nation’s history. There are many days before certification is performed on reported election results, so there is plenty of time to add this safeguard against potential AI-driven manipulation of the results.
Foreign countries have a strong incentive to turn AI hacking software loose on election machines in the United States. China already has AI so sophisticated that it generates realistic-looking movie scenes featuring famous Hollywood actors, like a dramatic fistfight between Brad Pitt and Tom Cruise on a rooftop that went viral.
The latest movie featuring Academy Award winner Casey Affleck is being filmed entirely with AI-generated scenes for the background. The actors and actresses are performing on an empty stage, saving many millions of dollars in production costs.
Oligarchs are making billions of dollars from AI, and yet some are running to federal politicians in Washington, D.C., to regulate them. So-called “doomers” are predicting the demise of humanity if the federal government does not step in to impose guardrails on what AI can do.
AI can well afford to create guardrails without the federal government coddling and protecting them. The federal government should take steps to prevent misuse of its weaponry by foreign regimes developing and using AI, which would not be subject to enforcement of any federal regulations anyway.
Unregulated competition among AI companies would be healthier than stifling federal intervention, and smaller AI players like Palantir oppose federal intervention. Deep State control of AI development from D.C. would be the worst of all scenarios.
John and Andy Schlafly are sons of Phyllis Schlafly (1924-2016) and lead the continuing Phyllis Schlafly Eagles organizations with writing and policy work.
