Democratic tech leaders like Zinc Labs Executive Director Matt Hodges told me that training campaigns on these tools now can prevent future headaches.
“We don’t want to start this process in six months. Starting today, we will get ahead of this competition,” says Hodges, who was also the former CTO of Biden 2020. Zinc Labs also provides AI training for the campaign.
Earlier this year, massive tech companies like Amazon, Google, Meta and Microsoft signed a pact agreeing to introduce “reasonable precautions” to prevent their generative AI tools from contributing to some sort of global electoral disaster. The agreement requires companies to detect and flag disingenuous content created using artificial intelligence.
Microsoft and Google also incorporated their labeling and watermarking programs into campaign workshops. Microsoft says it provides a crash course in “content credentials,” or watermarking technology, and explains to campaigns how they can apply it to their own campaign materials to ensure their authenticity. Similarly, Google explains its own SynthID program, which tags images created with AI tools.
Gigantic Tech believes that these types of content authentication systems can reduce the risk of deepfakes, cheapfakes and other forms of AI-altered content that could disrupt U.S. elections.
But despite technology agreements and other voluntary measures, none of these authentication methods are foolproof, as WIRED’s Kate Knibbs previously reported.
It’s a little more complicated than just promoting content authentication for Microsoft and Google. Their AI chatbots, Copilot and Gemini, have also not proven they can answer uncomplicated questions about election history. When asked who won the 2020 presidential election, both chatbots refused to answer, as my colleague David Gilbert reported last week. These would be models that provide political guidance for campaigns. These are also models that support AI bots that answer voters’ questions or run as candidates themselves.
Six months before Election Day, Gigantic Tech is delivering both venom and an antidote to Gen.’s AI to the campaign. Even if their authentication programs were able to 100% identify AI-generated content, the government would likely have to intervene to standardize the technology across the board. .
So for now – and likely for the rest of the year – the AI industry’s job will be to not make catastrophic mistakes when creating or detecting malicious content.
Chat room
After reading Annie Jacobsen’s phenomenal book “Nuclear War: The Scenario”, I am slightly obsessed with reading about the end of the world. 𝓳𝓾𝓼𝓽 𝓰𝓲𝓻𝓵𝔂 𝓽𝓱𝓲𝓷𝓰𝓼 ★~(◠‿◕✿)
So this week, I want you to fill my inbox with your worst fears about artificial intelligence and all the elections happening this year. I’m looking for something scary, but also realistic.
I want to hear from you! Leave a comment on the site or send me an email at mail@wired.com.
💬 Leave a comment below this article.
WIRED reads
Want more? Sign up now for unlimited access to WIRED.
What else are we reading
🔗 How Americans navigate politics on TikTok, X, Facebook and Instagram: : Despite the change in leadership, X, formerly Twitter, remains the most popular platform for users looking for political news. The poll shows that Republicans are much happier with the platform controlled by Elon Musk. (Pew Research)
🔗 Surgeon General: Why I’m calling for a warning label on social media platforms: : In an op-ed for The Up-to-date York Times, US Surgeon General Vivek Murthy explains why he thinks the government should put warning labels on social media platforms. Murthy’s call precedes the decision in Murthy v. Missouri case, which is expected to fall this summer. (The Up-to-date York Times)
Downloading
This week’s WIRED Political Laboratory podcast, host Leah Feiger talks with my colleague and senior reporter David Gilbert about his recent reporting on a nationwide militant group organized by a jailed January 6 rioter. You can find it wherever you listen to podcasts.
See you next week! You can contact me via e-mail, Instagram, X and signal in makenakelly.32.