Key points
- OpenAI's GPT 5.6-Sol discovered critical vulnerabilities in PHP during the week of October 19, 2026.
- A post-mortem revealed that a small team of maintainers knew of the issue two years before it was made public.
- At the time of the discovery, 73% of online PHP installations were running end-of-life versions with no active support.
The claim
According to Tom's Hardware, OpenAI's GPT 5.6-Sol identified critical vulnerabilities in the open-source programming language PHP during the week of October 19, 2026.reported A post-mortem published regarding the incident revealed that a small team of maintainers had known about the specific issue for two years prior to its public disclosure.An AI tool found a major security flaw in PHP that experts had known about for two years.GPT 5.6-Sol exposed a PHP remote code execution flaw rooted in longjmp() error handling known to maintainers for two years. The vulnerability stems from an ancient architectural decision in the PHP engine's error handling mechanism. When a PHP process runs out of memory, it calls the longjmp() C function to jump execution to another place. If this jump occurs while building a data structure, half-baked data remains in the memory heap. The cleanup mechanism then acts on these inconsistent leftover bytes instead of complete data.
Evidence: reported
PHP vulnerability post-mortem reveals bug was known two years ago — Small team of maintainers knew of the issue before it was made public, as new PHP2Shell chained exploit ravages serversaifleet-tomshardware
The response
Security experts noted that agentic AI exponentially sped up vulnerability scanning, rendering the usual 90-day disclosure window for security exploits pointless.reported The ongoing PHP2Shell exploit demonstrates how attackers can inject specific content into web requests to run any code with PHP's permissions.reported However, gaining a remote shell requires an attacker to know the exact version and build of PHP used, meaning it is not as trivial as simply pressing a hack button. The wide blast radius involves tens of millions of vulnerable servers, many of which are running the same exact version.
Evidence: reported
PHP vulnerability post-mortem reveals bug was known two years ago — Small team of maintainers knew of the issue before it was made public, as new PHP2Shell chained exploit ravages serversaifleet-tomshardware
Evidence: reported
PHP vulnerability post-mortem reveals bug was known two years ago — Small team of maintainers knew of the issue before it was made public, as new PHP2Shell chained exploit ravages serversaifleet-tomshardware
Why this matters now
PHP hosts around 70% of websites on the internet, making the remote code execution risk significant.reported The situation worsened because 73% of PHP installations online were running end-of-life versions with no active support at the time of the discovery.reported This combination of ubiquitous software and widespread use of unsupported versions elicited exasperation from those familiar with the ecosystem. The root cause was not just the vulnerability itself, but general complacency regarding mounting technical debt.
Evidence: reported
PHP vulnerability post-mortem reveals bug was known two years ago — Small team of maintainers knew of the issue before it was made public, as new PHP2Shell chained exploit ravages serversaifleet-tomshardware
Evidence: reported
PHP vulnerability post-mortem reveals bug was known two years ago — Small team of maintainers knew of the issue before it was made public, as new PHP2Shell chained exploit ravages serversaifleet-tomshardware
What buyers should weigh
Enterprise IT buyers must assess the proportion of their infrastructure running end-of-life PHP versions, as 73% of online installations lacked active support.reported The speed of AI-driven bug hunting means traditional disclosure windows may no longer provide adequate protection for unpatched systems. Vendor choice matters when considering platforms that rely on PHP with known architectural flaws in error handling. Organizations face risks from remote code execution when running unsupported versions of ubiquitous software packages.
Evidence: reported
PHP vulnerability post-mortem reveals bug was known two years ago — Small team of maintainers knew of the issue before it was made public, as new PHP2Shell chained exploit ravages serversaifleet-tomshardware
Why it matters
- AI agents can discover and exploit vulnerabilities faster than traditional human-led disclosure processes can manage.
- Running end-of-life software versions exposes organizations to remote code execution without any vendor support for fixes.
- Architectural technical debt in foundational technologies like PHP can create systemic risks across millions of servers.
- Knowledge of specific server versions becomes a critical factor in determining exploitability and risk levels.
Topics: PHP, OpenAI, GPT 5.6-Sol, vulnerability management, technical debt
Updated
Then / now / next
A small team of maintainers knew of a critical PHP architectural flaw two years before it was made public.
AI agents have discovered the vulnerability and exploits are actively targeting servers.
Organizations must determine if their 90-day disclosure windows remain viable against AI-driven scanning.
Story timeline
- Then A small team of maintainers knew of a critical PHP architectural flaw two years before it was made public.
- 04 Nov 2026 PHP vulnerability post-mortem reveals bug was known two years ago — Small team of maintainers knew of the issue before it was made public, as new PHP2Shell chained exploit ravages servers
- Now AI agents have discovered the vulnerability and exploits are actively targeting servers.
- Next Organizations must determine if their 90-day disclosure windows remain viable against AI-driven scanning.
The strongest doubt
Counterpoint from the sources
Gaining a remote shell requires knowing the exact version and build of PHP, so it is not as trivial as pressing a hack button.
For readers
Developers
- Understand that longjmp() usage during data structure building can leave inconsistent bytes in the memory heap.
- Ensure web requests are validated to prevent injection of content that exploits error handling mechanisms.
Business
- Assess the risk of 73% of online PHP installations lacking active support.
- Recognize that AI-driven bug hunting makes traditional disclosure windows less effective for risk management.
Policy
- Consider updating disclosure guidelines given that agentic AI can speed up vulnerability scanning exponentially.
- Evaluate the impact of maintaining end-of-life software versions in critical internet infrastructure.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
longjmp()
A C function that immediately jumps execution to another place without questions.
PHP2Shell
An exploit that allows attackers to run any code with PHP's permissions via web request injection.
Technical debt
Mounting issues in software architecture that accumulate over time and increase risk.
Questions readers ask
How does the PHP vulnerability allow remote code execution?
Exploiters inject specific content into web requests to manipulate inconsistent data left in the memory heap by the longjmp() function.
What percentage of PHP installations were unsupported?
According to the report, 73% of PHP installations online were running end-of-life versions with no active support.
Did maintainers know about the bug before it was public?
Yes, a post-mortem revealed that a small team of maintainers knew of the issue two years before it was made public.
Connected stories
Reading receipt
- OpenAI's GPT 5.6-Sol discovered critical vulnerabilities in PHP during the week of October 19, 2026.
- A post-mortem revealed that a small team of maintainers knew of the issue two years before it was made public.
- At the time of the discovery, 73% of online PHP installations were running end-of-life versions with no active support.
Still open: Will the 90-day disclosure window remain standard practice given the speed of AI bug hunting?
Still open: How many of the tens of millions of vulnerable servers will be patched before further exploitation?
