In October 2025, the Wayback Machine reached the milestone of one trillion archived web pages. Over 100,000 terabytes of data. This is a massive achievement for a nonprofit organization that has been operating since 1996. But behind this impressive number lies a difficult period that Internet Archive has gone through over the past year and a half. Cyberattacks, lawsuits, changes in access policies, and new challenges from AI companies ― all of this directly affects those who use the web archive for website restoration.

In this article, we'll break down what happened and what it means in practice.

The October 2024 Cyberattacks

In October 2024, the Internet Archive suffered a series of attacks that became the most severe in the organization's entire history.

On October 9, a pop-up message appeared on archive.org in which hackers announced the theft of the user database. As was later confirmed, the breach affected 31 million accounts ― email addresses, usernames, and password hashes. The 6.4 GB database was stolen through an unsecured authentication token in the organization's GitLab repository. This token had been left exposed for nearly two years.

Simultaneously with the data breach, archive.org was hit by a DDoS attack from the SN_BlackMeta group. The site was down for several days. On October 14, the Wayback Machine returned in read-only mode, and full functionality was only restored by the end of the month.

On October 20, yet another attack followed ― hackers gained access to the Zendesk support system through unrotated API tokens. Thousands of support tickets were compromised, including those to which users had attached personal documents. These tokens had not been changed even after the first breach, indicating serious problems with incident response.

Internet Archive founder Brewster Kahle assured that the archived data was safe, but the incident itself exposed the vulnerability of the organization's infrastructure, which operates on a minimal budget.

Lawsuits: Pressure from Publishers and Labels

Cyberattacks were not the only problem. Back in 2020, major publishers ― Hachette, HarperCollins, Penguin Random House, and Wiley ― filed a lawsuit against the Internet Archive over its digital book lending program, Open Library. In March 2023, the court ruled in favor of the publishers, and in September 2024, the appeals court upheld that decision. As a result, more than 500,000 books were removed from Open Library.

In parallel, major music labels ― Universal Music Group, Sony Music, and Concord ― filed a $621 million lawsuit over the Great 78 Project, in which the Internet Archive was digitizing old gramophone records. This case was settled in September 2025 on confidential terms.

These lawsuits don't directly affect the Wayback Machine or website restoration ― they concern books and music. But they create a serious financial burden on the organization and divert resources from its core activities. And any budget problems at the Internet Archive ultimately affect the stability and performance of all its services, including the Wayback Machine.

Publishers Blocking archive.org Crawlers

In 2025-2026, another alarming trend emerged. Major news publishers began restricting Internet Archive crawlers' access to their websites.

The New York Times completely blocked archive.org crawlers and added archive.org_bot to its robots.txt. The Guardian restricted access to article pages, leaving only homepage and section pages available in the Wayback Machine. The Financial Times blocks all external bots, including Internet Archive crawlers.

The reason is concern that AI companies use data from the Wayback Machine to train language models. Publishers believe that the Wayback Machine API can serve as a convenient access point to their content for machine learning systems. And these concerns are not unfounded: an analysis of the Google C4 dataset, used to train the T5 and Llama models, showed that web.archive.org was among the top 200 most represented domains in the training data.

Recently an AI company was sending tens of thousands of requests per second to Internet Archive servers, which led to a temporary service outage. Such incidents became one of the reasons publishers started reconsidering their relationship with the web archive.

For website restoration, this is not yet critical ― the Wayback Machine continues to archive the vast majority of the internet. But if the trend of blocking archive.org crawlers continues, gaps in the archives will grow, especially in content from major media outlets. This means that restoring a website that linked to materials from such publications will become more difficult.

Stricter Rate Limits and Download Blocks

The Internet Archive has always rate-limited requests to its API, but after the events of 2024, these restrictions have become tighter. The CDX API allows an average of 60 requests per minute. When the limit is exceeded, the server responds with a 429 (Too Many Requests) code. If the client continues to ignore 429 responses for more than a minute, the IP address is blocked at the firewall level for one hour. Each subsequent violation doubles the blocking time.

In practice, this means that downloading a large website from the Wayback Machine from a single IP address has become significantly slower and riskier. Many third-party scripts and utilities for downloading from the web archive don't account for these limitations and lead to users getting blocked.

Our system has adapted to these changes. To download data from the Wayback Machine, we use multiple proxy servers, which allows us to distribute the load and stay within the limits. This ensures stable operation even when downloading large websites with hundreds of thousands of pages, without the risk of being blocked and without having to wait for hours due to rate limiting.

AI and Restored Content: A New Challenge

Artificial intelligence is changing not only how we search for information but the internet itself. And this directly concerns website restoration from archives.

The first problem is AI content in archives. Starting around 2023, the internet has seen an enormous amount of text generated by language models. The Wayback Machine archives everything indiscriminately, without distinguishing whether text was written by a human or a machine. If you're restoring a website whose snapshots were taken after 2023, there's a chance that some of the content had already been replaced by the owners with AI-generated text. This is especially relevant for websites that changed ownership or went through periods of neglect.

The second problem is that search engines are changing their approach to AI content. Google is actively fighting low-quality AI texts, demoting them in search results. If a restored website contains such content, it may face indexing problems. When restoring a website, it's worth checking the content for typical signs of machine generation and, if necessary, rewriting or removing such texts.

The third problem is AI-generated search results. The Internet Archive is already experimenting with archiving responses from ChatGPT and AI summaries in Google search results. This changes the very concept of what it means to "save a web page." A page used to be a static document; now it can contain dynamically generated AI content that differs from query to query.

On the other hand, AI also opens up positive possibilities. Language models can be used for automatic cleanup of restored content: fixing broken markup, removing ad blocks and navigation elements, restoring text structure, even converting outdated templates to modern formats. But that's a topic for a separate article.

Archiving JavaScript Websites: An Unsolved Problem

There's another issue that has existed for a while but is becoming more acute every year. Modern websites are increasingly built on JavaScript frameworks ― React, Vue, Angular. Content on such websites is generated dynamically in the browser, and when a crawler accesses the server, it receives an empty HTML template.

The Wayback Machine can save JavaScript files, but it doesn't always correctly reproduce dynamically generated pages. The more complex the framework and the more a website depends on external APIs, the worse the archiving result.

In practice, this means that websites built as SPAs (Single Page Applications) are archived less effectively than classic HTML websites. And if the trend toward JavaScript-heavy frameworks continues, the proportion of "properly recoverable" websites in the archive will gradually decline.

What This All Means in Practice

Despite all the challenges, the Wayback Machine remains the primary and irreplaceable source of archived web page copies. One trillion saved pages is a colossal volume of data, and for most website restoration tasks, this data is more than sufficient.

But relying exclusively on archive.org is becoming riskier than before. Here's what to keep in mind:

The data is available now ― but there's no guarantee it will be available tomorrow. Publishers are blocking crawlers, limits are tightening, and financial pressure on the organization is growing. If you're planning a restoration, don't postpone it.

When restoring websites with snapshots taken after 2023, check the content for AI-generated texts. Especially if the website changed owners or topics.

JavaScript-heavy websites (SPAs built with React, Vue, Angular) may be archived incompletely. For such websites, snapshots from earlier periods, when the site still used classic server-side rendering, may be of better quality.

Don't rely on a single source. Check alternative archives and search engine caches. Sometimes the version of a website you need can be found where the Wayback Machine didn't save it.

The use of article materials is allowed only if the link to the source is posted: https://archivarix.com/en/blog/webarchive-2026/