There is concern about the rise in misleading generative AI-based political ads, which is causing more confusion as the political season heats up. In turn, this will make it easier to sell lies.
The rising temperature of political ads between former President Donald Trump and Florida Governor Ron DeSantis for the 2024 GOP nomination makes this actual November 2019 photo out of place today. Could this be a Generative AI Image? No, this is the original picture.
The FEC recently voted to open public comment on how to regulate generative AI political ads. But news organizations need to learn how to report on these ads. This is their opportunity.
The DeSantis PAC attack ad on frontrunner Donald Trump is a recent example. In it, the ad featured the former president criticizing Republican Governor Kim Reynolds of Iowa, except that Trump’s voice was generated, but Trump did not speak and said the words. PAC supporting DeSantis copied text and produced audio from Trump’s text post his voice, In July, WESH 2 News (a TV outlet in Central Florida) did an effective debunking.
This is a great start and news outlets can do much more. The key question is: how does generative AI transform the already complex issue of political ads, deception and lies? Is it exactly the same, just formulated differently and more sophisticated, or are there fundamentally new questions along with responsibilities for news organizations to think about?
This is the unique power of generative AI used in political ads. It destroys the authenticity that we currently only associate with the whole person. Any element of a person’s likeness can now be created separately by AI, especially for public figures. It could generate voice or image or video independently of the generated text or language and a combination of the above.
Journalists and editors now need to think about the authenticity of equality in the age of generic AI.
Generative AI can extract your likeness as a complete person – voice, visual, appearance, gestures, eye movements, smile and what not. Given this, the various normative elements of any given public figure can be synthesized and recombined arbitrarily.
Until recently, the focus has been on the content of the claims in an advertisement. There are also ongoing standardization efforts to help encode the origin and authenticity of branded videos and prevent tampering. But, as in a political ad the authenticity of your likeness can itself be stripped and re-attached for the obvious purpose of propaganda and deception, this generative AI capabilities brings up new issues.
What should the newsroom do?
I call the Journalism and Media Ethics Council of the Markkula Center. Several council members addressed this topic during one of our meetings. One of my learnings: Ask key questions related to the content – that is, fact-check the claims. But also dig in and uncover the digital mechanics in advertising.
Questions journalists can ask when considering reporting advertising are:
- Who issued the advertisement? Is the ad about the candidate the ad sponsor supports, or does it contain messages against or attacking an opponent? If the ad is on behalf of a candidate and contains lies or falsehoods about the candidate, the FEC may look into it. If the ad implicates a competitor, and lies about them, the FEC has no immunity.
- Is this sarcasm? For example, The Daily Show created a satirical generic AI video on Biden, which is easy to tell because of his branding on the scene. Lincoln Project added a laugh track In a video President Trump posted about the impeachments and the committee on January 6th.
- If it is not satire, then the demand for review becomes high. Identify where Generative AI was used. Video, audio, text, titles, a mix of all? Has the manufacturer disclosed this?
- Dig around and find out if the public figure in the ad actually said these things somewhere in the speech. Details matter.
- Even if they did say so, that doesn’t automatically make the claim right or wrong. A claim or claim review (also called fact-checking) will be required. Or if it was already fact-checked, those results – such as true, false, partially true, etc., would already be available on major fact-checking organizations or peer news sites.
- Is advertising using generative AI to create a likeness of a person in order to disseminate an otherwise true or factual message? The newsroom needs to know if producers are trying to cut costs by creating a likeness of the candidate instead of recording a traditional new video. This raises the question of whether the candidate has authorized equality.
- Similarly, is advertising using generative AI to create a likeness of a person simply to broadcast a rhetorical claim, and therefore not fact-checkable because it is too general? Such claims are protected political speech anyway. For example, Donald Trump may never have actually said that “Biden messed up badly with his handling of Maui”, but a generic AI ad could simply show him saying so using Trump’s likeness. In what case did he authorize that rendering?
As you can tell, based on the answers to these questions, reporters and editors will be in a better position to decide whether and how they want to report the claims.
What is the real opportunity?
There is a real opportunity to go beyond the content of the ads. News outlets can use generic AI ads to educate the public about how generic AI works and how it can be used.
When? disguise 2 DeSantis reported the attack ad, it The video was labeled “Generative AI Political AD” while the reporter was discussing the mechanics. Additional labeling won’t hurt at all. A news report on the ad might add a second line to the running label:
“Advertisement includes audio generated using text posted by Person X or Y on Platform Z. Person X did not say that.”
This gives the viewer a clear idea – whether the viewer is listening to the reporter or not – and goes beyond the usual umbrella label “Generative AI Political AD”.
Going forward, Generative AI Ads are inputs for great explainers:
In addition to quick revelations, journalists can write explanatory articles on generative AI ads to give the public an idea of how the sausage was made.
Use the seven questions above as a lecturer checklist. Help people see how AI works in ads, as distinct from other forms of advertising creativity. Explain the concepts of training data and language models. Debunk the myths about what the term artificial intelligence really means. Machines do not understand meaning like humans.
This will help in generative AI literacy at a time when it is most needed as the 2024 election cycle is in full swing. The more background knowledge we have as political ad viewers, the more likely the ad could pique our curiosity about the generic AI being used in it. In turn, we may seek clarification, and this will slow us down before believing the claims or sharing the ad too quickly online. This extreme irrational behavior is exactly what deceptive and manipulative campaigns seek.
There are no magic bullets, no one answer fits all. Debunk, explain, educate.