Opinion | AI makes it easier to deny reality. Michigan politics are already suffering

The Nixon tapes would not end Richard Nixon’s presidency today.

In 1974, the decisive evidence in Watergate was not a rumor or a partisan claim. It was a tape. When the Supreme Court forced Nixon to release his Oval Office recordings in United States v. Nixon, Americans accepted a basic premise: the tape was real. The fight was over what it meant, not whether it existed.

Keith Kindred taught at the high school and college levels for 33 years and now writes about history, education, and public policy. (Courtesy photo)

Now, in the age of AI, the assumption that evidence is what it appears to be is fraying. Our courts and politics still depend on people agreeing on basic facts, and that agreement is much weaker now than it was during most of my 33-year teaching career.

Consider Warren, Michigan. In 2018, longtime mayor Jim Fouts was recorded making derogatory remarks about women and Black residents. He claimed the recordings had been manipulated. Despite the controversy, he won re-election. Even by that time, the lesson was already emerging: authentic evidence no longer automatically compels agreement.

Republican candidates in Michigan, including state Senator Aric Nesbitt and businessman Perry Johnson, have already utilized hyper-realistic AI-generated video parodies during campaigns. Although framed as satire, these AI-generated clips are so convincing that they challenge the authenticity of all political media and make it easy for voters to be misled.

Generative AI did not create deception, but it has made convincing fakes far cheaper to produce. Legal scholars Bobby Chesney and Danielle Citron call this the “Liar’s Dividend”. Once people know that audio and video can be fabricated, the existence of fakes becomes a ready-made defense against the real thing. A public figure no longer needs to prove a recording is false. They only need to suggest that it could be.

For centuries, new technologies, from printing to photography to DNA analysis, generally made it easier to verify what happened. Generative AI is different in scale. It does not just spread information faster; it weakens the idea that proof is settled by what people can see and hear.

That shift is no longer theoretical. In cases tied to the Jan. 6 attack on the Capitol, defense lawyers have already raised the possibility that video evidence could be manipulated. As synthetic media improves, those arguments will become more common and more plausible to juries and the public.

This struggle is visibly playing out across Lansing, as Michigan’s state government attempts to build guardrails for a technology moving faster than its bureaucracy. The Department of Education recently took the unusual step of issuing statewide AI guidance for K-12 schools to manage its sudden presence in classrooms. Meanwhile, state Rep. Jaime Greene has proposed creating a new Artificial Intelligence Governing Board to oversee and vet how state agencies use generative tools. These reactive, piecemeal measures from schools and lawmakers illustrate that trying to police the consumption of AI after the fact is a losing battle.

The fix cannot just be external oversight; it has to be built into the digital media system itself. We need widely adopted provenance standards that verify media at the moment it is created. Efforts such as the Coalition for Content Provenance and Authenticity are developing systems that attach tamper-evident metadata to images, audio and video. Those records can show when a file was created, what device produced it, and whether it has been altered.

Think of it as a chain of custody for digital media. Just as courts require a documented chain of custody or a notarized deed to trust evidence, digital media would require verified origin.

Still, technology alone will not solve the problem. These standards should be backed by incentives and rules. Courts should increasingly condition admissibility on verifiable provenance. News organizations should prioritize authenticated material. Platforms should clearly signal or down-rank content that lacks verification. Lawmakers should establish penalties for deliberately stripping or falsifying provenance data.

None of this will eliminate deception. Determined actors will still find ways to mislead, but it can change the default. Today, any piece of evidence can be doubted. In a provenance-based system, unverified media should be treated with caution from the start.

Richard Nixon was undone because Americans still agreed on what a tape was. Cases like Jim Fouts’s show how quickly that agreement can fray. Generative AI is accelerating that erosion, giving public figures a powerful new tool to evade accountability.

The question is not whether fake media will exist. It already does. The question is whether authentic evidence will still matter.