In 2023, during a presentation that circulated widely through technology policy circles, Center for Humane Technology co-founder Tristan Harris and technologist Aza Raskin delivered a stark assessment of artificial intelligence and democracy. In their talk, Raskin suggested that the 2024 election cycle could become the “last human elec tion.” The phrase was less prediction than warning. Harris and Raskin argued that generative AI systems—capable of producing persuasive text, images, audio, and video on demand—were arriving faster than democratic institutions could adapt.

The idea did not come from the political class. It emerged from technologists watching the acceleration of machine-generated media. In the presentation, Harris and Raskin explained how rapidly improving generative models could scale persuasion beyond anything produced by traditional campaigns, advertising firms, or media organizations. Synthetic political messaging, once difficult to create, could soon be generated in unlimited variations.

Three years later, that warning is resurfacing in a different environment.

Artificial intelligence companies that once described themselves as research labs are now deeply embedded in global infrastructure—from search engines to cloud computing systems and government technology programs. Among them is the company behind ChatGPT.

OpenAI began in 2015 as a nonprofit research initiative founded by Sam Altman, Elon Musk, and several leading AI researchers. The organization’s founding mission was to build advanced artificial intelligence in a way that would benefit humanity broadly rather than concentrate power within a handful of companies.

But the cost of training large AI models quickly forced a structural shift. In 2019 OpenAI created a capped-profit subsidiary designed to raise the billions of dollars needed for computing infrastructure while maintaining oversight from its nonprofit board.

That transition transformed the company from a research initiative into one of the most influential AI firms in the world. Microsoft invested heavily and integrated OpenAI systems into its cloud platform and software ecosystem.

The technology now powers tools used daily by hundreds of millions of people.

At the same time, OpenAI’s systems are moving into government environments.

Earlier this year the company confirmed an agreement allowing its models to operate within secure government networks used by the U.S. Department of Defense. The Pentagon has been expanding artificial intelligence programs for years as part of broader modernization efforts, according to the Defense Department’s.

The partnership enables government agencies to test generative AI for tasks such as software development, data analysis, and operational planning.

On its own, the contract does not involve elections or political messaging. But it represents a broader shift underway across the technology sector.

Companies building the most advanced generative models are becoming national security contractors.

That development brings the Harris-Raskin warning into sharper focus.

Artificial intelligence systems capable of generating persuasive content are now embedded inside institutions that also manage intelligence systems, cybersecurity operations, and digital infrastructure.

The overlap is structural, not conspiratorial.

Political campaigns already rely heavily on data analytics, targeted messaging, and digital advertising. Generative AI expands those capabilities dramatically. A single model can produce thousands of variations of political messaging tailored to different audiences.

Researchers studying election security have warned that synthetic media and automated persuasion tools could accelerate misinformation campaigns if used irresponsibly.

OpenAI itself has acknowledged those risks. Ahead of the 2024 global election cycle, the company announced safeguards intended to prevent its systems from producing targeted political persuasion or election interference content, according to reporting from.

Still, the broader AI industry remains divided about how closely companies should work with governments.

Anthropic, a competing AI company founded by former OpenAI researchers, has emphasized AI safety and regulatory oversight in its public policy work. The company has supported stronger testing and evaluation requirements for advanced models before widespread deployment.

In policy discussions around national security and AI governance, Anthropic has also pushed for transparency and regulatory clarity regarding how powerful AI systems should be classified and controlled.

Those disagreements reveal a deeper tension inside the AI sector.

One side argues that rapid development is necessary to maintain technological leadership in an increasingly competitive global landscape. Another camp warns that deployment is moving faster than democratic institutions can build safeguards.

The Harris and Raskin presentation explored precisely that imbalance.

The talk described how generative AI could flood digital networks with persuasive synthetic content. Unlike earlier technologies—radio, television, or social media—AI systems can create political messaging instantly and endlessly.

Campaign slogans, speeches, videos, and images can be generated automatically and distributed across social networks.

For election administrators and researchers, the concern is not only misinformation. It is scale.

When persuasive messaging can be produced automatically, the volume of political communication could expand beyond the capacity of traditional oversight or fact-checking systems.

Sam Altman has acknowledged similar risks in public testimony and policy discussions. He has repeatedly argued that advanced AI will require new regulatory frameworks to address its societal impact.

The United States now approaches the 2026 midterm elections during a period of extraordinary technological change.

Generative AI tools are integrated into everyday software. Governments are experimenting with AI systems across defense and intelligence programs. Political campaigns are beginning to explore automation and data-driven communication strategies.

None of those developments guarantee that elections will be manipulated by machines.

But they illustrate the transformation Harris and Raskin were describing.

For most of modern history, political persuasion depended on human infrastructure—speechwriters, strategists, media organizations, and campaign staff.

Artificial intelligence introduces a new actor into that system: machines capable of producing persuasive communication at industrial scale.

The technology does not vote.

Yet the systems shaping public information increasingly rely on algorithms designed and controlled by a small number of companies.

The phrase “last human election” therefore functions less as a prophecy than as a question.

If democratic institutions cannot keep pace with the technologies shaping public discourse, future elections may still be human contests—but they will take place inside an information ecosystem increasingly influenced by machines.