# The software that powers 70% of the web has just one employee

> The PHP Foundation sustains the language that powers up to 70% of the web with one full-time employee, 14 part-time contributors and donations.
> By José Maldonado · Cybersecurity · 2026-10-29

In the week when two serious PHP flaws forced half the internet to patch urgently, the foundation that sustains the language was explaining how it works on the inside. [Elizabeth Barron](https://www.rappler.com/technology/php-foundation-under-funded-understaffed/), its executive director, summed it up in one sentence: there is one full-time person, who is her, and fourteen collaborators, mostly part-time. The board is unpaid and the money comes from donations and grants.

PHP may be behind as much as 70% of the world's websites, according to estimates used in the industry. That figure coexists with a structure that fits in a small room, and the contrast is not an anecdote: it is a description of the model on which much of the web has been built.

The comparison with this week's incident is inevitable. The patches released on Wednesday came from a team that works with external researchers and with companies such as OpenAI, Wiz and Calif, and that coordinates the response from a non-profit organization. We covered this when analyzing the [php2shell flaw and its impact on the development chain](https://observatorioblockchain.com/en/cybersecurity/php2shell-php-rce-critico/)the technical problem was serious, and the question that remains is who pays for the next one.

## An "unacceptable" template

With one full-time person and fourteen part-timers, the foundation cannot afford the luxury of reviewing everything that comes in. Its role is to coordinate, fund those who investigate and maintain the project's infrastructure, while the code is written by volunteers who work at other companies.

That model works when there are few proposals and breaks down when volume grows. Barron has explained that the project receives almost twice as many proposed code changes as a year ago, and that the number of contributors has not doubled. Reviewing those contributions requires time from experienced people, which is the scarcest resource.

The second consequence is dependence on third parties. When a critical flaw appears, the foundation works with researchers and companies that provide analysis and testing capacity. It is a useful network and also a reminder that the language's security depends on those companies continuing to have an interest in collaborating.

## The double-edged sword of artificial intelligence

This week's episode leaves an interesting data point about the role of AI in open source software. The two flaws had been in the code for a decade and were found with the help of automated tools, which make it possible to review volumes of code that were previously unmanageable.

That same advance generates work. If anyone can generate a patch with a model, the number of proposals rises, and each one needs human review so as not to introduce a bigger problem. The foundation has described it without embellishment: more changes, same reviewers.

It is a pattern that repeats in other projects. We saw it in the analysis of the [25% of proposals generated by AI in open source software](https://observatorioblockchain.com/en/ai/ia-open-source-25-por-ciento-pull-requests-slop/), where the conclusion was that the bottleneck has shifted from writing code to reviewing it. AI makes the first part cheaper and does not touch the second.

Resource

Situation described

Full-time staff

1 person (the executive director)

Contributors

14, mostly part-time

Board

Unpaid volunteers

Funding

Donations and grants

Estimated footprint

Up to 70% of websites

Code proposals

Almost double that of a year ago

## Why does nobody pay for PHP (and open source software in general)?

Open source software has a well-known incentive problem. The companies that benefit most from a language are the ones that pay least to maintain it, because the benefit is diffuse and the cost is concrete. Each company prefers that someone else pay.

In the case of PHP, the aggravating factor is that the language underpins products with large revenues. E-commerce platforms, content managers, internal banking systems and public administrations run on PHP code. None of those users appears on the usual list of donors.

The comparison with physical infrastructure helps. Nobody expects a bridge to be maintained with voluntary donations, because the cost of its collapse is obvious and falls on those who use it. In software, that cost is spread across millions of people, and that is why nobody feels responsible.

## A week that puts the focus where it belongs

The two recently published flaws forced international coordination within hours. Cloudflare reported millions of attack attempts, and the foundation has said it has no evidence that they were exploited before disclosure or in systems that had already been patched.

That outcome is better than could have been expected, and it should not be too reassuring. The response worked because there were researchers willing to act, companies that provided data and a foundation that coordinated. If any of those pieces had been missing, the story would be different.

The other reading concerns the timeline. The flaws had been in the code for a decade, which means the problem was not new: it was invisible. AI accelerated discovery and, with it, the need to respond quickly.

## What it costs to maintain a language

A programming language is not just an interpreter. It is documentation, tests, support for old versions, vulnerability management and a community that answers questions. All of that consumes time, and the time of people with judgment is what no small donation can buy.

When funding is insufficient, the effects do not appear as an immediate collapse. They appear as decisions that are postponed: a review that becomes more superficial, a patch that takes longer, a proposal that goes unanswered. Over time, that turns into security debt.

The foundation has said that PHP is "incredibly underfunded" for the dependence placed on it. It is an uncomfortable sentence for the person saying it, and it is also the most honest conclusion published this week.

## Companies must contribute to PHP

The first step is to know how much they depend on the language. Many organizations do not have an inventory of their own infrastructure and discover that half of their critical systems run on PHP when there is an incident.

The second is to fund it. A company that saves an entire team thanks to a language should contribute a fraction of that saving to the project that maintains it, whether with money, engineering time or infrastructure.

The third is to take part in review. Contributing code is useful, and contributing quality reviews is more so, because the bottleneck is in evaluating what others propose.

None of the three solves the underlying problem, which is structural. And all three together turn an invisible dependency into a relationship with obligations, which is the first step so that the next decade-old flaw is found sooner and fixed with more people than are available today.
