The same companies that broke social media now control AI.

2004 · The Promise

"Connect the world."

Facebook launches from a Harvard dorm room. The mission is to bring people together. The founders believe it. So does everyone else.

VS

2022 · The Promise

"Benefit all of humanity."

OpenAI launches ChatGPT. The mission is to ensure artificial general intelligence benefits all of humanity. The founders believe it. So does everyone else.

2008 · The Race

Engagement becomes the metric.

Whatever holds attention wins. Whatever provokes outrage gets reach. The algorithm doesn't care what's true. It cares what gets clicked.

VS

2023 · The Race

Capability becomes the metric.

Whatever scores higher on benchmarks wins funding. Whatever ships first wins market share. Safety teams are rebranded. Then they are quietly dissolved.

2014 · The First Crack

Cambridge Analytica scrapes 87 million users.

Without consent. Without compensation. The data is used to target political ads. The platform shrugs. Years pass before anyone is held accountable. No one ever really is.

VS

2024 · The First Crack

AI models train on the entire scraped internet.

Every book. Every blog post. Every Reddit comment. Every line of code. Every photograph. Without consent. Without compensation. The platform shrugs. Lawsuits pile up. The training continues.

2017 · The Children

Internal Facebook research finds Instagram makes teen girls feel worse about themselves.

Teen suicide rates begin a vertical climb. The research is buried. When it leaks, the company says it has been "mischaracterized." The product does not change.

VS

2023 · The Children

Snapchat's AI cheerfully advises a 13-year-old how to lie to her parents about meeting a 31-year-old man.

"They went from homework assistant to suicide assistant in the course of six months."— Tristan Harris

2021 · The Whistleblowers

Frances Haugen leaks 10,000 documents.

She testifies before Congress. There are speeches. There are hearings. There are headlines for a week. Nothing changes. The product does not change.

VS

2023–25 · The Whistleblowers

Hinton leaves Google. Gawdat: "We f---ed up." Hitzig leaves OpenAI: "reform must come from outside."

There are speeches. There are hearings. There are headlines for a week. Nothing changes. The product does not change.

2018, 2019, 2020, 2021, 2022 · The Apologies

Mark Zuckerberg apologizes to Congress.

Then again. Then again. Then again. Then again. Engineers bolt new safety features onto an unchanged business model. The metric is still engagement. The product does not change.

VS

2023–26 · The Apologies

Sam Altman tours world capitals warning of existential risk.

Then he ships GPT-4. Then GPT-5. xAI accelerates. Anthropic accelerates. Google DeepMind accelerates. Meta accelerates. The metric is still capability. The product does not change.

"That fairly simple technology was enough in the first contact with AI to break humanity — with information overload, addiction, doom scrolling, sexualization of kids, polarization, fake news, and breakdown of democracy."

— Tristan Harris, The AI Dilemma

Now imagine the same incentive structure with technology one thousand times more powerful.

The Legal Math

Section 230 shielded social media from liability for what users posted.

For 25 years, the platforms were not legally responsible for what appeared on them. That law is the only reason the modern internet exists in the form it does.

— Communications Decency Act, 1996

VS

The Legal Math

Section 230 does not apply to AI.

"The platform is providing the information in a different kind of way. It's not that they're just a neutral vessel for information." When AI causes harm, there is no legal shield. There is also, currently, no legal framework.

— Zoe Hitzig, former OpenAI governance researcher

"Nukes don't make stronger nukes. AI makes stronger AI."

— Tristan Harris