← Home

The lag

One story on a single timeline. What was built, when the harm was first documented, and when the law finally responded. Read top to bottom.

Every entry links to the page on this site that carries its source. The timeline adds no new claims of its own — it is a map of the argument made everywhere else.

How to read this page

Read it top to bottom. The left track is what the industry built. The middle track is when each harm was first documented — in peer-reviewed research, official investigations, or court. The right track is when the law responded. Every entry links to the page on this site that carries its evidence.

The shape of the page is the argument: the Laws column stays almost empty while the other two fill up.
1995
Symptom: Stormfront, the first major hate website, is set up — a decade before social media. source →
1996
The Machine: DoubleClick begins tracking people across sites with cookies. source →
Law: US Section 230 shields online services from liability for what users post. The one early law — and it protects the intermediary rather than giving the person harmed a remedy. source →
1998
The Machine: Google is founded. source →
Symptom: A retracted MMR–autism paper is published, seeding decades of anti-vaccine content. source →
2000
The Machine: Google AdWords launches: ads sold against the words people search for. source →
2002
The Machine: Google starts selling its search ads by auction, priced per click. source →
2004
The Machine: Facebook launches. source →
Symptom: The OSCE documents online hate feeding real-world hate crime. source →
2007
The Machine: The iPhone goes on sale — “the Internet in your pocket”, in Apple’s words; Facebook opens advertising. source →
2012
The Machine: Facebook buys Instagram, the photo-sharing app built for the phone. source →
Symptom: A decline in adolescent mental health is later dated to around this year, mainly on US surveys; how widespread it was, and what caused it, are contested. source →
2013
Symptom: Researchers link girls’ Facebook photo activity to body-image concern; whether the feed causes it is still contested. source →
2016
Symptom: Election interference at scale is documented. source →
2018
The Machine: Facebook re-weights the feed for “meaningful social interactions” — ranking for reactions and comments becomes the default. source →
Symptom: Cambridge Analytica: it emerges that data from up to 87 million Facebook profiles, harvested without consent, went to the company for its US political work. source →
Symptom: UN investigators link Facebook to violence against the Rohingya in Myanmar. source →
Law: GDPR becomes enforceable across the EU. source →
↑ first privacy law with penalties on the companies' own scale — 23 years after the first documented harm
2020
Symptom: Christchurch: New Zealand's Royal Commission documents the attacker's online activity. source →
Symptom: The COVID “infodemic”. source →
Symptom: Reported losses to scams that start on social media begin an eight-fold climb (2020–2025). source →
2021
Symptom: The Facebook Files: Frances Haugen discloses the company’s own research into the harm it had measured. source →
2022
The Machine: Generative AI (ChatGPT) is expected to make personalised manipulation cheap. source →
Law: The EU Digital Services Act is passed. source →
2023
Law: The UK Online Safety Act becomes law, putting duties of care on the services themselves. source →
2024
The Machine: AI Overviews; the rise of zero-click search. source →
Symptom: Police report “violence as a service” recruiting children online. source →
Law: The DSA applies; Australia legislates a minimum age of 16 for social-media accounts. source →
Law: The EU AI Act enters into force; its bans on manipulative and exploitative AI systems apply from February 2025. source →
2025
Law: The EU issues its first DSA fine (X, €120 million). source →
2026
Law: A California bellwether verdict in the social-media harms litigation; Meta settles with 51 US attorneys general for up to $17.1 billion, with time limits for under-18s. source →

Why the lag matters

What the layout shows is a delay. The response track stays almost empty for two decades, and its single early entry — Section 230, in 1996 — shielded the intermediary rather than giving anyone harmed a remedy. Hate sites were documented in 1995; earlier laws touched the edges, but the first EU-wide rules with penalties large enough to matter — the GDPR for data in 2018, the DSA for platform design in 2024 — came decades later. Election interference was documented in 2016; the rules written to address it, the Digital Services Act, applied in 2024. The gap is not the same as malice. It is what it looks like when technology moves at the speed of code and law moves at the speed of institutions.

A note on how this page is built

This is a narrative timeline — one form of what Segel and Heer call narrative visualization — a format one 2026 study (a CHI conference paper on scrolling privacy policies) found keeps comprehension intact while making a dense subject easier to follow. The three tracks run side by side, in parallel lanes; each track is marked by shape and colour so it works without colour, and the content is plain HTML that reads on a slow connection and with JavaScript switched off. The design choices follow established accessibility guidance.

Sources

Did this page help?