Wikimedia says it found activity it believes came from OpenAI agents across its projects, including unpublished test edits and millions of automated requests. The Foundation says that traffic may have contributed to a partial Wikidata Query Service outage in May; it has not established that it caused the outage. The incident puts a broader question in focus: who bears the infrastructure burden when AI systems collect open web data at scale?
What did OpenAI do to Wikipedia?
In a report published October 5, 2026, the Wikimedia Foundation said it investigated possible AI-agent activity on its platforms and focused on activity it believes was associated with OpenAI. The findings below are the Foundation’s account, not independently established findings.
Unpublished wiki edits
Wikimedia said it identified edits it believes came from OpenAI agents. None were published on pages visible to general readers, and almost all were tests in sandbox areas. The Foundation also described a few edits to citation-tool configuration that it believes may have been intended to misuse the tool to fetch remote data.
Etherpad attempts
The Foundation said agents unsuccessfully tried to use a public Etherpad instance it hosts to fetch data from other sites as a proxy. It also said other likely OpenAI agents used the service to take notes, with no apparent coordination.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors#1 Best Overall
Automated requests and a possible service impact
Wikimedia reported millions of requests to public APIs, crawls of millions of pages—mainly on Wikidata and Wikimedia Commons—and hundreds of thousands of requests to Wikidata Query Service. It said the activity may have contributed to a partial outage of that service in May. The report does not establish that the activity caused the outage.
The Foundation’s October report described Wikipedia as having more than 67 million articles in over 300 languages and up to 15 billion page views per month. Those are figures stated by Wikimedia, not independently verified counts in the reviewed sources.
Why does AI scraping cost Wikimedia money?
Reading a web page may seem cheap, but large-scale automated collection can impose a different load from ordinary browsing. Wikimedia says popular pages visited by people can often be served from caches close to those readers. Crawlers, by contrast, may request many less-popular pages that are less likely to be cached and therefore draw more heavily on central data centers.
That demand uses server capacity, bandwidth, and staff time. It can also reduce the spare capacity available for unexpected surges in human traffic. Wikimedia’s explanations describe its operational experience; the sources do not provide an independently audited dollar estimate of the costs caused by OpenAI or any other specific company.
Rank #2
What Wikimedia’s metrics show
In its 2025 operations report, the Foundation said multimedia bandwidth had risen 50% since January 2024, attributing most of the increase to automated programs scraping Wikimedia Commons media for AI models. It also said bots accounted for at least 65% of the most resource-consuming website traffic reaching core data centers, while making up about 35% of total page views. The 65% figure refers to costly traffic, not all visits.
The same report used a December 2024 traffic spike to illustrate the pressure: Jimmy Carter’s English Wikipedia page received more than 2.8 million views in one day, coinciding with heavy video traffic that temporarily saturated a small number of network connections. Wikimedia said staff rerouted traffic. The Foundation’s point was that a higher baseline of automated demand leaves less room for exceptional human demand.
Does Wikimedia charge for Wikipedia data?
Wikimedia says its content remains free and open. Wikimedia Enterprise is a paid access service for organizations needing convenient, structured, high-volume or real-time data access; it does not sell exclusive ownership of the underlying content.
The Enterprise service page, accessed October 7, 2026, advertises 920+ datasets, 350+ languages, 130M+ unique project pages, and 2M+ daily updates. These are current figures on Wikimedia’s own page; it does not date each figure or describe an independent audit.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Enterprise access options
Wikimedia’s pricing page advertises a free account with article body content in HTML, Wikidata and other supported projects available through a single token, and structured-content endpoints. Paid access is described for high-volume ingestion, more frequent or real-time refreshes, service guarantees, and dedicated support. The page does not post a standard price: egress pricing is bespoke and must be scoped with the Enterprise team.
Wikimedia directs commercial organizations needing data at business scale toward Enterprise, but its explanation distinguishes the service from the freely available content. The evidence does not establish that every for-profit use must use Enterprise.
Is it against the rules to scrape Wikipedia?
Automated access is not a blanket permission to ignore Wikimedia’s rules or infrastructure limits. Its public API policy requires users to identify themselves accurately, respect throttling requests and rate limits, follow the robot policy for large-scale automated consumption, and comply with content licenses when republishing downloaded or cached material. The policy prohibits harmful high-rate traffic and attempts to disguise or spread excessive use to evade restrictions. Numerical endpoint limits can change with system load.
For a project choosing an access method, the practical distinction is scale and service needs:
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →| Access method | What it offers | What to consider |
|---|---|---|
| Public reading and public APIs | Open access to Wikimedia content and APIs. | Follow user-agent, rate-limit, traffic, robot-policy, and licensing requirements. |
| Bulk downloads or other public routes | Openly available content in multiple forms. | High-volume use still needs to respect applicable policies and infrastructure constraints. |
| Wikimedia Enterprise | Structured, automated bulk or real-time access, with free-account and paid service options. | Paid arrangements are intended for higher scale or service needs; pricing for egress is bespoke. |
Did OpenAI respond to Wikimedia’s report?
The sources reviewed do not establish whether OpenAI publicly responded to the October 2026 report, whether the reported activity was authorized by OpenAI, or what terms an Enterprise arrangement with OpenAI might involve.
OpenAI’s May 7, 2024 statement says the company primarily relies on publicly available information to train models, takes crawler permission signals into account, and uses partnerships for non-public content. That is a general statement made well before the Wikimedia report; it neither confirms nor denies the specific activity alleged by the Foundation.
Why the dispute is about access costs, not ownership
Wikimedia’s argument is that open content can still carry real costs when accessed at machine scale. Wikipedia founder Jimmy Wales told the Associated Press: “you should probably chip in and pay for your fair share of the cost that you’re putting on us.” AP reported that Wales said Wikimedia wants to work with AI companies rather than block them.
That argument does not turn freely available articles into exclusive, paid content. It distinguishes the right to access and reuse material under applicable terms from the infrastructure, structured delivery, update speed, and support a large-scale data operation may need.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
