Based on reporting byCNBCFrance 24Al Jazeera

PHP security flaws trigger worldwide site outages: what six AIs agree on, and what they added or left out

11 min readTechnology
PHPHPCLCloudflareOPOpenAIWIWizCACalifNSNSACybersecurityArtificial intelligenceSoftware vulnerabilitiesPHP2ShellPHP vulnerabilityWebsite security
Satcove NewsTechnology6 AI models, 3 reportsAutomated analysis, not reviewed by a person
PHP source code on a screen
Illustrative photo: PHP source code on a screen, not from the reported event. lmonk72, CC0, via Wikimedia Commons.

Quick answer

  • Reported: CNBC, France 24, and Al Jazeera documented government websites in Europe and elsewhere going offline Wednesday after PHP vulnerabilities were disclosed.
  • Stated by PHP Foundation: Al Jazeera reported the foundation called the flaws among the most severe it had encountered and urged website administrators to update immediately.
  • Stated by NSA: Al Jazeera reported the NSA advised taking websites offline if they could not be updated immediately.
  • Panel result: no direct contradiction, but 4 critical and 4 neutral omissions across 12 points.
  • Reliability signal: GPT and Grok added 2 claims the cited reports do not support.

2 / 12

Points all models share

10 / 12

Points where they differ

0

Direct contradictions

6 of 6

Models that ran

Who said what

Same reports, independent analyses. Highlighted rows are the points where the models that ran differ.

PointClaudeGPTGeminiMistralPerplexityGrok
AI tools discovered two critical vulnerabilities in PHP, a widely used programming language.Includes the pointIncludes the pointIncludes the pointIncludes the pointIncludes the pointIncludes the point
PHP powers approximately 70-75% of websites whose programming language can be identified.Neutral omissionIncludes the pointMentions it only in general termsLeaves it outIncludes the pointLeaves it outIncludes the point
The vulnerabilities allow attackers to execute remote commands without authentication on vulnerable servers.Includes the pointIncludes the pointIncludes the pointIncludes the pointIncludes the pointIncludes the point
PHP Foundation or security firms called the vulnerabilities among the most severe in PHP's history.Neutral omissionLeaves it outLeaves it outLeaves it outIncludes the pointLeaves it outIncludes the point
A PHP security patch was released on Wednesday at approximately 1345 GMT.Critical omissionIncludes the pointMentions it only in general termsIncludes the pointLeaves it outIncludes the pointLeaves it out
Government websites in multiple European countries went offline as a precautionary measure.Different emphasisIncludes the pointMentions it only in general termsIncludes the pointIncludes the pointIncludes the pointIncludes the point
The NSA advised organizations to patch PHP immediately or take vulnerable systems offline.Critical omissionLeaves it outIncludes the pointIncludes the pointIncludes the pointIncludes the pointIncludes the point
Cloudflare observed millions of exploitation attempts following public disclosure of the vulnerabilities.Different emphasisIncludes the pointMentions it only in general termsIncludes the pointIncludes the pointIncludes the pointIncludes the point
Security firm Calif discovered the vulnerabilities using OpenAI's GPT-5.6 Sol model in August.Neutral omissionIncludes the pointLeaves it outIncludes the pointLeaves it outLeaves it outLeaves it out
Wiz security firm independently discovered vulnerabilities weeks after Calif using AI models including Google's Gemini.Neutral omissionIncludes the pointLeaves it outIncludes the pointLeaves it outLeaves it outLeaves it out
Hackers were actively exploiting the vulnerabilities before the security fix was released.Critical omissionIncludes the pointLeaves it outLeaves it outLeaves it outLeaves it outLeaves it out
Vulnerability details leaked a day before the planned patch release, prompting pre-emptive shutdowns.Critical omissionLeaves it outLeaves it outLeaves it outLeaves it outLeaves it outIncludes the point

What each mark means

  • ✓Includes the point
  • ~Mentions it only in general terms
  • 0Leaves it out
  • –Model did not run on this story

How a difference is classified

  • ContradictionTwo models assert incompatible things.
  • Critical omissionA model leaves out a point that changes how a reader understands the story.
  • Neutral omissionA model leaves out a minor detail.
  • Different emphasisAll models cover it, but with different weight.
  • Unverified additionA claim the cited reports do not support.

The panel, model by model

ClaudeClaude

Stayed in the reports
Thesis
Critical PHP vulnerabilities discovered via AI prompted worldwide website shutdowns and emergency patching across government and corporate sectors.
Only this model
Hackers were actively exploiting the vulnerabilities before the security fix was released.
Left out
PHP Foundation or security firms called the vulnerabilities among the most severe in PHP's history.; The NSA advised organizations to patch PHP immediately or take vulnerable systems offline.
In its words
“Billions of websites and critical infrastructure rely on PHP.”

GPTGPT

Added unverified claims
Thesis
Critical vulnerabilities in PHP led to widespread website shutdowns as organizations rushed to secure their systems against potential attacks.
Only this model
Nothing that no other model also said.
Left out
PHP Foundation or security firms called the vulnerabilities among the most severe in PHP's history.; Security firm Calif discovered the vulnerabilities using OpenAI's GPT-5.6 Sol model in August.
In its words
“This incident raises significant concerns about the security of a major portion of the internet, highlighting vulnerabilities that could lead to widespread cyberattacks, disrupting services globally.”

GeminiGemini

Stayed in the reports
Thesis
Critical security vulnerabilities in the widely used PHP programming language, uncovered by artificial intelligence, prompted mass website outages and emergency patches globally as governments and corporations rushed to prevent cyberattacks.
Only this model
Nothing that no other model also said.
Left out
PHP powers approximately 70-75% of websites whose programming language can be identified.; PHP Foundation or security firms called the vulnerabilities among the most severe in PHP's history.
In its words
“The discoveries highlight a dangerous new phase where AI accelerates the discovery of severe software vulnerabilities, threatening internet-scale infrastructure and forcing a massive, urgent global patching effort across governments, corporations, and public services.”

MistralMistral

Stayed in the reports
Thesis
Critical PHP vulnerabilities discovered via AI led to global website outages as governments and corporations rushed to patch systems.
Only this model
Nothing that no other model also said.
Left out
A PHP security patch was released on Wednesday at approximately 1345 GMT.; Security firm Calif discovered the vulnerabilities using OpenAI's GPT-5.6 Sol model in August.
In its words
“The vulnerabilities pose a massive risk to global cybersecurity, potentially exposing millions of servers to remote attacks.”

PerplexityPerplexity

Stayed in the reports
Thesis
AI-assisted discovery of two critical PHP vulnerabilities triggered precautionary website shutdowns worldwide before a patch became available, exposing the difficulty of rapidly securing internet-scale infrastructure.
Only this model
Nothing that no other model also said.
Left out
PHP powers approximately 70-75% of websites whose programming language can be identified.; PHP Foundation or security firms called the vulnerabilities among the most severe in PHP's history.
In its words
“Unpatched PHP servers could permit unauthenticated remote command execution, potentially enabling theft, website modification, or broader disruption.”

GrokGrok

Added unverified claims
Thesis
AI-assisted discovery of two critical PHP flaws triggered worldwide site outages and urgent patching to prevent remote server takeovers.
Only this model
Vulnerability details leaked a day before the planned patch release, prompting pre-emptive shutdowns.
Left out
A PHP security patch was released on Wednesday at approximately 1345 GMT.; Security firm Calif discovered the vulnerabilities using OpenAI's GPT-5.6 Sol model in August.
In its words
“Millions of servers running PHP could be exposed to unauthenticated remote code execution, risking data theft, defacement or ransomware at internet scale.”

Verified timeline

DateEventSourceCertainty
August 11, 2026Security firm Calif reported the first PHP vulnerability discovered using OpenAI's GPT-5.6 Sol modelAl JazeeraReported by outlets
August 24, 2026Wiz security firm reported the second vulnerabilityAl JazeeraReported by outlets
Tuesday (October 21)Vulnerability details leaked online, a day before planned patch releaseFrance 24Reported by outlets
Wednesday, approximately 1345 GMTPHP developers released a new version fixing the vulnerabilitiesCNBCReported by outlets
Wednesday (October 22)Government websites across Europe went offline as precautionary measure; NSA advised organizations to patch immediately or take systems offlineAl JazeeraStated by officials or the organisation concerned
Late WednesdayCloudflare CEO reported millions of exploitation attempts after public disclosure, jumping from thousands before announcement to millions within hoursFrance 24Reported by outlets

The sources differ on the estimated percentage of websites using PHP: CNBC cites roughly 75 percent, while France 24 and Al Jazeera cite about 70 percent.

Why this matters

PHP powers approximately 70-75% of websites whose programming language can be identified, according to CNBC. The two critical vulnerabilities discovered through AI allow attackers to "remotely run commands on a server running old PHP code" without authentication, according to Al Jazeera. France 24 reported that Cloudflare saw millions of exploitation attempts after public disclosure, with thousands of attempts occurring before the announcement and jumping to hundreds of thousands within two hours.

Government websites across Europe and Asia went offline as precaution. CNBC reported sites in the Netherlands and Finland went dark, while France 24 documented shutdowns in France, Belgium, and the Czech Republic. The U.S. National Security Agency advised organizations to patch immediately or take vulnerable systems offline, according to Al Jazeera.

Small websites on shared hosting face particular risk. CNBC noted such sites lack technical resources to patch quickly and may remain exposed longer after fixes become available. The PHP Foundation warned that AI tools now find security bugs faster than humans can patch them.

What is established, alleged and stated

Reported by the outlets

  • According to CNBC, PHP developers identified two critical vulnerabilities and released a fix on Wednesday at approximately 1345 GMT, with reports that hackers were exploiting the flaws before the patch became available.
  • France 24 reports that vulnerability details leaked online Tuesday, a day before the planned fix release, prompting government agencies to take websites offline as a precautionary measure.
  • According to Al Jazeera, the two vulnerabilities allow attackers to remotely run commands on vulnerable servers without password or account access.
  • CNBC reports that Cloudflare observed "early attacks emerging" and was updating firewall rules, while France 24 states that Cloudflare saw millions of exploitation attempts within hours of public disclosure.

Stated by officials or the organisation concerned

  • According to France 24, the PHP Foundation stated the vulnerabilities were "among the most severe vulnerabilities discovered in PHP, and arguably in software history."
  • CNBC reports Automattic stated it was "working with the PHP security team to ensure a smooth update."

Alleged or disputed

Nothing in the sources.

What we still do not know

  • CNBC reports that officials said "all PHP versions from 5 onward could be affected," but PHP had not yet disclosed which versions are actually vulnerable.

  • The sources differ on when vulnerability details became public. France 24 states details leaked Tuesday, one day before the planned patch release, but CNBC does not mention this leak.

  • The reports do not specify how many websites were successfully compromised versus merely targeted by exploitation attempts.

  • CNBC notes the vulnerability "could potentially open the door to millions of servers" but does not estimate actual breach numbers.

  • The sources do not establish when government and corporate websites taken offline as precautions returned to service.

  • Al Jazeera states the flaws "had been present for decades" but does not explain why they remained undiscovered until AI tools identified them.

What the models agree on

  • AI tools discovered two critical vulnerabilities in PHP, a widely used programming language.
  • The vulnerabilities allow attackers to execute remote commands without authentication on vulnerable servers.

Agreement between models does not validate an allegation. It only shows that they read the same reports in the same way.

Where they differ

Every difference is typed. An omission is not treated as an error: it is only called critical when leaving the point out would change how a reader understands the story.

DivergenceTypeWhy it mattersModels concerned
A PHP security patch was released on Wednesday at approximately 1345 GMT.Critical omissionEstablishes when the fix became available to organizations.Left out by Mistral and Grok
The NSA advised organizations to patch PHP immediately or take vulnerable systems offline.Critical omissionEstablishes official U.S. government response and guidance to the vulnerability threat.Left out by Claude
Hackers were actively exploiting the vulnerabilities before the security fix was released.Critical omissionIndicates attackers had access to exploit the vulnerabilities during the window before patching.Left out by GPT, Gemini, Mistral, Perplexity and Grok
Vulnerability details leaked a day before the planned patch release, prompting pre-emptive shutdowns.Critical omissionExplains why governments took websites offline before the official patch release date.Left out by Claude, GPT, Gemini, Mistral and Perplexity
Government websites in multiple European countries went offline as a precautionary measure.Different emphasisDemonstrates the widespread impact of the vulnerabilities on critical infrastructure.GPT (general terms only)
Cloudflare observed millions of exploitation attempts following public disclosure of the vulnerabilities.Different emphasisDemonstrates active exploitation of the vulnerabilities after details became public.GPT (general terms only)
PHP powers approximately 70-75% of websites whose programming language can be identified.Neutral omissionProvides context on PHP's prevalence and therefore the scale of potential impact.Left out by Gemini and Perplexity
PHP Foundation or security firms called the vulnerabilities among the most severe in PHP's history.Neutral omissionEmphasizes the severity assessment from authoritative sources in the PHP ecosystem.Left out by Claude, GPT, Gemini and Perplexity
Security firm Calif discovered the vulnerabilities using OpenAI's GPT-5.6 Sol model in August.Neutral omissionIdentifies the initial discoverer and specific AI model used in the discovery process.Left out by GPT, Mistral, Perplexity and Grok
Wiz security firm independently discovered vulnerabilities weeks after Calif using AI models including Google's Gemini.Neutral omissionDocuments that multiple security firms independently discovered the same vulnerabilities.Left out by GPT, Mistral, Perplexity and Grok

Model reliability scorecard

ModelCoverageCritical omissionsUnverified claimsContradictionsEditorial reliability
Claude75%200Solid
GPT58%210Weak
Gemini67%200Solid
Mistral58%300Mixed
Perplexity50%200Mixed
Grok67%210Mixed

How the rating works. Coverage is the share of the panel's points that a model includes. Each model was asked for 6 to 10 claims, so coverage reflects what it chose to include within that budget. Penalty points: 1 per critical omission, 2 per unverified claim, 2 per contradiction, 1 if coverage is below 60%. Strong: coverage of 80% or more and no penalty. Solid: coverage of 65% or more and at most 2 penalty points. Mixed: at most 4 penalty points. Weak: more than 4. There is no numerical grade.

The most important finding

The main result is not that the models disagree. None of them contradicted another. They differ mostly by omission: 4 critical and 4 neutral across 12 points, plus 2 differences of emphasis. The more important signal is reliability: GPT and Grok added 2 claims that the cited reports do not support, and those matter more than any omission.

Reading between the lines

What this suggests: Al Jazeera reports that the underlying flaws "had been present for decades" yet required "a chain of steps at the edge of what human researchers could practically achieve" to exploit. This implies AI's contribution was not discovering a new weakness but demonstrating exploitability at scale. If this reading holds, it would mean the vulnerabilities existed within reach of sophisticated human attackers for years before AI exposure—a fact that would support urgency around patch adoption going forward.

What this does not establish: CNBC and France 24 confirm that Cloudflare saw "millions" of exploitation attempts after disclosure, but neither source specifies how many attacks succeeded in compromising systems or extracting data. Active attempts and successful breaches are distinct events the reporting does not resolve.

The key signal: France 24 notes that vulnerability details "leaked online Tuesday, a day before the planned release of the fix," forcing premature shutdowns. This timing gap suggests that rapid disclosure windows—whether accidental or engineered—may now constrain defenders' ability to patch methodically, amplifying pressure on organizations with limited technical capacity.

What to watch

  1. Will small websites and shared hosting providers patch quickly enough to prevent widespread exploitation? CNBC reported that small sites on shared hosting services could "take longer to patch and remain exposed well after fixes become available," creating a window for attackers.

  2. How many exploitation attempts will occur now that vulnerability details are public? France 24 reported that Cloudflare CEO Matthew Prince observed "millions" of exploitation attempts within hours of public disclosure, after "thousands" in the 10 hours before announcement.

  3. Which PHP versions will PHP officially confirm as vulnerable? CNBC noted that officials said "all PHP versions from 5 onward could be affected," but PHP had not yet disclosed which versions are actually vulnerable.

  4. Will the PHP Foundation's warning about AI-accelerated vulnerability discovery prompt coordinated industry responses? Al Jazeera reported the foundation called for "a coordinated response to avoid major disruptions" as AI tools find security weaknesses faster than human programmers.

  5. How quickly will government websites taken offline return to service? France 24 reported websites in multiple countries went dark as a precautionary measure, leaving unclear timelines for restoration after patching.

Planned review of this page: October 29, 2026.

Method, honestly stated

  • Every model received the same reports (CNBC, France 24, Al Jazeera) and the same prompt, and was told to use only those reports.
  • Models that ran: Claude, GPT, Gemini, Mistral, Perplexity and Grok. No model was left out.
  • Each claim a model made was checked against the reports. A claim counts as verified if its supporting passage appears word for word in a report, or if a separate model judged it a supported paraphrase. Otherwise it is marked unverified.
  • A separate model merged equivalent claims into comparable points and classified each difference as contradiction, critical omission, neutral omission, different emphasis or unverified addition. That step is automated and can be wrong.
  • The similarity score (91/100) measures how alike the analyses read. It is not an accuracy or confidence score, not agreement on facts and not a measure of truth.
  • An automated pipeline wrote and published this page. No person read it before publication.

Verdict

Coverage of the central fact is broadly consistent: 2 of 12 points appear in every analysis. The differences are mostly about completeness and framing, not contradiction. The main risk identified is added claims: GPT and Grok included material the cited reports do not support.

What we checked

Each claim a model added was searched for in the full text of the cited reports.

ClaimSaid byResult
Cloudflare observed millions of exploitation attempts following public disclosure of the vulnerabilities.ClaudeSupported paraphrase
Cybersecurity firms used AI to uncover two critical vulnerabilities in PHP, affecting about 70% of the world's websites.GPTSupported paraphrase
Cybersecurity experts emphasized the risks posed by AI-assisted hacking, leading to increased attack attempts post-announcement.GPTUnverified
The NSA urged organizations using PHP to update immediately or take vulnerable sites offline.GPTSupported paraphrase
Cloudflare saw exploitation attempts jump to millions following the public disclosure.GeminiSupported paraphrase
PHP vulnerabilities affect roughly 70% of websites worldwide.MistralSupported paraphrase
Cloudflare reported millions of exploitation attempts following the vulnerability disclosure.MistralSupported paraphrase
The PHP Foundation called the flaws among the most severe in PHP's history.MistralSupported paraphrase
Cloudflare reported that exploitation attempts increased from thousands to millions after public disclosure.PerplexitySupported paraphrase
The reports differ on PHP’s affected scope, with one saying all versions from PHP 5 onward could be affected and another limiting exposure to self-hosted PHP systems.PerplexitySupported paraphrase
52 model claims checked: 41 matched word for word in the reports, 9 judged supported paraphrases, 2 not supported by the reports.All modelsIn the reports

Frequently asked

What vulnerabilities were discovered?#
France 24 and Al Jazeera report two critical vulnerabilities in PHP programming language. Al Jazeera states these allow attackers to remotely run commands on servers without password or account access.
How did AI contribute to finding these flaws?#
France 24 reports security firm Calif found flaws in August using OpenAI's GPT-5.6 Sol model. Al Jazeera states AI tools enabled researchers to test more attacks than human programmers could alone.
Why did government websites go offline?#
CNBC reports government agencies in Netherlands and Finland took sites offline ahead of fix availability as precaution. France 24 notes vulnerability details leaked online Tuesday before planned Wednesday fix release.
What percentage of websites use PHP?#
CNBC cites web technology tracker W3Techs reporting roughly 75% of websites with identifiable programming language run PHP. France 24 and Al Jazeera report approximately 70%.
Who discovered the PHP vulnerabilities?#
Al Jazeera reports that security firm Calif reported the first flaw on August 11 using OpenAI's GPT-5.6 Sol model, and that Wiz reported the second on August 24 using AI models including Google's Gemini 4 Argon.
How many exploit attempts did Cloudflare see?#
France 24 reports Cloudflare saw thousands of attempts in the 10 hours before the announcement, tens of thousands within 45 minutes after it, hundreds of thousands within 2 hours, and then "millions". The figure comes from Cloudflare's CEO, Matthew Prince, and is not independently verified.

The full analysis from each model

Each model's structured answer, shown in full and unedited. Run October 22, 2026, same prompt for every model.

Claude

Thesis

Critical PHP vulnerabilities discovered via AI prompted worldwide website shutdowns and emergency patching across government and corporate sectors.

Claims (10)

  1. Two critical vulnerabilities in PHP programming language were identified and publicly disclosed.CNBC“Developers of the PHP programming language said Tuesday they had identified two critical vulnerabilities”
  2. PHP powers approximately 70-75% of websites whose programming language can be identified.France 24“PHP code helps run about 70 percent of the world's websites”
  3. Artificial intelligence models were used to discover the PHP vulnerabilities.Al Jazeera“Cybersecurity experts discovered the pair of vulnerabilities, dubbed PHP2Shell, with the help of artificial intelligence (AI) models.”
  4. A PHP security fix was released on Wednesday at approximately 1345 GMT.CNBC“They said around 1345 GMT on Wednesday that a new version fixing the vulnerabilities was live.”
  5. Government websites in multiple European countries went offline as a precautionary measure.CNBC“government agencies in the Netherlands and Finland were among those that took sites offline ahead of the fix as a precaution.”
  6. Hackers were actively exploiting the vulnerabilities before the fix was released.CNBC“with reports that hackers were exploiting the flaws to break into websites”
  7. Cloudflare observed millions of exploitation attempts following public disclosure of the vulnerabilities.France 24“Cloud services firm Cloudflare saw millions of exploitation attempts after the public disclosure of PHP's critical vulnerabilities”
  8. Security firm Calif discovered vulnerabilities using OpenAI's GPT-5.6 Sol model in August.France 24“Security firm Calif found the flaws in August using OpenAI's GPT-5.6 Sol model.”
  9. Wiz security firm independently discovered vulnerabilities weeks after Calif using AI models including Google's Gemini 4 Argon.France 24“Wiz, which is owned by Google, said it found them independently weeks later with help from AI models including Google's Gemini 4 Argon”
  10. The vulnerabilities allowed attackers to execute remote commands without authentication on vulnerable PHP servers.Al Jazeera“an attacker can remotely run commands on a server running old PHP code over the internet, without needing a password or legitimate account access.”

Left open

  • Which specific PHP versions are vulnerable - SOURCE 1 states officials said versions 5 onward could be affected but PHP had not disclosed which versions are actually vulnerable
  • The exact timeline of when details leaked online - SOURCE 2 states details leaked Tuesday, a day before the planned fix release, but SOURCE 1 does not mention this leak
  • Whether Calif and Wiz discovered the same vulnerabilities or different ones - SOURCE 2 suggests they were independent discoveries while SOURCE 3 implies they are the same pair of vulnerabilities

Stakes

Billions of websites and critical infrastructure rely on PHP. The vulnerabilities could enable attackers to compromise servers at scale without authentication. AI-accelerated exploitation threatens millions of systems, particularly small websites lacking technical resources. Extended patching windows leave prolonged exposure for shared hosting customers.

What this model would watch

  • Adoption rate of PHP patches across website operators, particularly small and shared hosting sites
  • Number and sophistication of active exploitation attempts as attackers weaponize the vulnerabilities
  • Timeline for restoring government and corporate websites taken offline as precaution
  • Whether AI-discovered vulnerabilities accelerate the pace of future security flaws in other widely-used software
GPT

Thesis

Critical vulnerabilities in PHP led to widespread website shutdowns as organizations rushed to secure their systems against potential attacks.

Claims (8)

  1. A significant number of websites, including government sites, were taken offline globally as a precaution against vulnerabilities in PHP.CNBC“government websites in Europe and elsewhere went offline”
  2. Cybersecurity firms used AI to uncover two critical vulnerabilities in PHP, affecting about 70% of the world's websites.France 24“the flaws rank among the most severe vulnerabilities discovered in PHP”
  3. The vulnerabilities allow attackers to run commands on servers running vulnerable PHP code remotely, without authorization.Al Jazeera“mean an attacker can remotely run commands on a server running old PHP code”
  4. PHP programmers released a patch to fix the vulnerabilities on Wednesday, following their discovery earlier this year.France 24“PHP programmers on Wednesday released a patch fixing the vulnerability”
  5. Cybersecurity experts emphasized the risks posed by AI-assisted hacking, leading to increased attack attempts post-announcement.France 24“Cloudflare saw millions of exploitation attempts after the public disclosure”
  6. The NSA urged organizations using PHP to update immediately or take vulnerable sites offline.Al Jazeera“the United States National Security Agency (NSA) advised taking websites offline”
  7. Shutdowns affected various sectors, including government, transport, and media, highlighting widespread disruption due to PHP vulnerabilities.CNBC“several organizations followed the example of the National Cyber Security Center and took public-facing websites offline”
  8. The vulnerabilities were discovered with the aid of AI models, marking a new phase in vulnerability detection.Al Jazeera“AI tools let researchers find vulnerabilities in software by testing more attacks than human programmers can alone”

Left open

  • There is some discrepancy in the percentage of websites affected, as one source cites 70% and another mentions 75%.
  • It is unclear how many specific websites in different countries were affected during this event.

Stakes

This incident raises significant concerns about the security of a major portion of the internet, highlighting vulnerabilities that could lead to widespread cyberattacks, disrupting services globally.

What this model would watch

  • The response from tech companies regarding the patch deployment and security measures.
  • Potential rise in cyberattacks utilizing these vulnerabilities before patches are universally applied.
  • Further developments from AI-assisted vulnerability discoveries and subsequent impacts on cybersecurity practices.
Gemini

Thesis

Critical security vulnerabilities in the widely used PHP programming language, uncovered by artificial intelligence, prompted mass website outages and emergency patches globally as governments and corporations rushed to prevent cyberattacks.

Claims (8)

  1. Critical vulnerabilities in the PHP programming language were discovered with the help of artificial intelligence.Al Jazeera“AI was used to pinpoint two vulnerabilities in the PHP language”
  2. PHP programmers released a patch fixing the vulnerability on Wednesday.France 24“PHP programmers on Wednesday released a patch fixing the vulnerability.”
  3. Government agencies and companies around the world shut down websites as a precaution ahead of or during the fix.CNBC“government agencies in the Netherlands and Finland were among those that took sites offline ahead of the fix as a precaution.”
  4. The United States National Security Agency advised organizations to patch their sites immediately or pull vulnerable systems offline.France 24“the US National Security Agency told organisations to patch their sites immediately or pull vulnerable systems offline.”
  5. Security firm Calif found the flaws using OpenAI's GPT-5.6 Sol model as part of an audit.France 24“Security firm Calif found the flaws in August using OpenAI’s GPT-5.6 Sol model.”
  6. Wiz found the second vulnerability independently weeks later with help from AI models.France 24“Wiz, which is owned by Google, said it found them independently weeks later with help from AI models”
  7. Cloudflare saw exploitation attempts jump to millions following the public disclosure.France 24“That jumped to tens of thousands ​to exploit it. That jumped to tens of thousands ​in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now”
  8. The vulnerabilities could allow attackers to run commands on servers without a username or password.France 24“attackers to get into your servers without any authorisation, without any username or password”

Left open

  • The exact percentage of global websites run by PHP, with sources varying between about 70 percent and roughly 75 percent.
  • The precise timeline of when details leaked online versus when public disclosures officially occurred.

Stakes

The discoveries highlight a dangerous new phase where AI accelerates the discovery of severe software vulnerabilities, threatening internet-scale infrastructure and forcing a massive, urgent global patching effort across governments, corporations, and public services.

What this model would watch

  • The speed and extent of global patch adoption by small websites and shared hosting providers.
  • Further statements or metrics from cloud security firms regarding ongoing exploit attempts.
  • Additional security disclosures or policy responses driven by AI-assisted vulnerability findings.
Mistral

Thesis

Critical PHP vulnerabilities discovered via AI led to global website outages as governments and corporations rushed to patch systems.

Claims (8)

  1. AI tools discovered two critical vulnerabilities in PHP, a widely used programming language.France 24“Security firm Calif found the flaws in August using OpenAI’s GPT-5.6 Sol model.”
  2. PHP vulnerabilities affect roughly 70% of websites worldwide.France 24“Wiz estimated that PHP code helps run about 70 percent of the world's websites.”
  3. Government websites in multiple countries went offline as a precaution.CNBC“Government agencies in the Netherlands and Finland were among those that took sites offline.”
  4. The NSA advised organizations to patch PHP immediately or take systems offline.Al Jazeera“Effective Immediately, all organizations using PHP for their systems are advised to be prepared for an update.”
  5. Cloudflare reported millions of exploitation attempts following the vulnerability disclosure.France 24“Cloudflare saw millions of exploitation attempts after the public disclosure.”
  6. The vulnerabilities allow remote code execution without authentication.Al Jazeera“An attacker can remotely run commands on a server running old PHP code over the internet, without needing a password.”
  7. AI-assisted discovery of vulnerabilities is accelerating cybersecurity risks.France 24“Increasingly powerful AI models can spot weak points in computer scripts that have gone unnoticed by humans.”
  8. The PHP Foundation called the flaws among the most severe in PHP's history.Al Jazeera“The group said the flaws were among the most severe it had encountered.”

Left open

  • Which specific PHP versions are affected by the vulnerabilities.
  • The exact number of websites or services that remain unpatched.
  • Whether the vulnerabilities were exploited before the patch was released.
  • The full extent of data breaches or unauthorized access resulting from the flaws.

Stakes

The vulnerabilities pose a massive risk to global cybersecurity, potentially exposing millions of servers to remote attacks. AI-driven discovery of such flaws could lead to more frequent and severe cyber threats, requiring urgent and coordinated responses from governments and corporations.

What this model would watch

  • The rate of PHP patch adoption across different sectors and regions.
  • Reports of successful exploits or data breaches linked to the vulnerabilities.
  • Government and corporate responses to future AI-discovered vulnerabilities.
  • The impact of AI tools on both offensive and defensive cybersecurity strategies.
Perplexity

Thesis

AI-assisted discovery of two critical PHP vulnerabilities triggered precautionary website shutdowns worldwide before a patch became available, exposing the difficulty of rapidly securing internet-scale infrastructure.

Claims (8)

  1. PHP developers released a fix for two critical vulnerabilities on Wednesday.CNBC“a new version fixing the vulnerabilities was live”
  2. Government agencies in the Netherlands and Finland took websites offline as a precaution.CNBC“government agencies in the Netherlands and Finland were among those that took sites offline”
  3. The vulnerabilities could allow attackers to remotely execute commands without authentication.Al Jazeera“an attacker can remotely run commands on a server running old PHP code over the internet”
  4. The flaws were discovered with assistance from artificial-intelligence models.Al Jazeera“Cybersecurity experts discovered the pair of vulnerabilities, dubbed PHP2Shell, with the help of artificial intelligence (AI) models.”
  5. The NSA advised organizations to update PHP immediately or take vulnerable systems offline.France 24“the US National Security Agency told organisations to patch their sites immediately or pull vulnerable systems offline”
  6. Cloudflare reported that exploitation attempts increased from thousands to millions after public disclosure.France 24“That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now”
  7. France temporarily made local-government administrative websites inaccessible while addressing the vulnerabilities.CNBC“France's Interior Ministry said it had temporarily taken local government administrative websites offline”
  8. The reports differ on PHP’s affected scope, with one saying all versions from PHP 5 onward could be affected and another limiting exposure to self-hosted PHP systems.CNBC“all PHP versions from 5 onward could be affected”

Left open

  • The reports do not identify the exact vulnerable PHP versions.
  • They do not establish how many websites were successfully compromised, rather than merely targeted.
  • They differ on the estimated share of websites using PHP, citing roughly 70 percent versus roughly 75 percent of identifiable-language sites.
  • They do not specify when every temporarily disabled website returned to service.
  • The reports do not fully reconcile the discovery timeline, including the roles of Calif, Wiz, OpenAI, and other AI models.

Stakes

Unpatched PHP servers could permit unauthenticated remote command execution, potentially enabling theft, website modification, or broader disruption. The scale of PHP use and rapid AI-assisted vulnerability discovery increase pressure on organizations with limited patching capacity.

What this model would watch

  • Whether additional exploitation or confirmed compromises are reported.
  • Which PHP versions are officially confirmed as vulnerable.
  • How quickly governments, companies, hosting providers, and small-site operators apply the patch.
  • Whether further websites or online services experience outages during remediation
Grok

Thesis

AI-assisted discovery of two critical PHP flaws triggered worldwide site outages and urgent patching to prevent remote server takeovers.

Claims (10)

  1. Two critical PHP vulnerabilities were found with AI tools.France 24“Security firm Calif found the flaws in August using OpenAI’s GPT-5.6 Sol model.”
  2. Wiz called the flaws among the most severe in PHP and arguably in software history.France 24“It called the flaws "among the most severe vulnerabilities discovered in PHP, and arguably in software history".”
  3. The NSA advised organizations to patch PHP immediately or take systems offline.Al Jazeera““Effective Immediately, all organizations using PHP for their systems are advised to be prepared for an update addressing two critical vulnerabilities,” the NSA said.”
  4. The Dutch National Cyber Security Center took its own sites offline as a precaution.CNBC“The Dutch cybersecurity agency itself went offline, saying it had done so as a precaution because of the vulnerabilities.”
  5. Cloudflare observed millions of exploitation attempts within hours of the disclosure.France 24“That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.”
  6. PHP powers roughly 70-75 % of websites whose programming language is identifiable.CNBC“PHP runs roughly 75% of websites whose programming language can be identified, according to web technology tracker W3Techs.”
  7. France’s Interior Ministry temporarily deactivated local-government administrative websites.CNBC“France's Interior Ministry said it had temporarily taken local government administrative websites offline while awaiting a security fix.”
  8. The PHP Foundation said the flaws were among the most severe it had encountered in over a decade.France 24“Volker Dusch, a security engineer at the foundation, told the Wall Street Journal it was the most serious problem PHP had faced in more than a decade.”
  9. The flaws allow remote command execution on servers without any authentication.Al Jazeera“The two loopholes mean an attacker can remotely run commands on a server running old PHP code over the internet, without needing a password or legitimate account access.”
  10. Details leaked a day before the planned patch release, prompting pre-emptive shutdowns.France 24“Details leaked online Tuesday, a day before the planned release of the fix, prompting the rush to take sites offline before hackers could exploit them.”

Left open

  • Which exact PHP versions are vulnerable
  • How many sites remained offline after the patch became available
  • Whether AI-assisted attacks have already succeeded against unpatched servers

Stakes

Millions of servers running PHP could be exposed to unauthenticated remote code execution, risking data theft, defacement or ransomware at internet scale.

What this model would watch

  • Evidence of successful exploitation in the wild
  • Volume of sites still running unpatched PHP versions
  • Additional AI-discovered vulnerabilities in other widely used software
  • Coordinated patching response by major hosting and CMS providers
The exact prompt

Instructions

You are an independent news analyst. Below are reports of the same event from several outlets. Use ONLY these reports: do not add outside facts and do not use prior knowledge.

Required answer format (JSON, no other text)

thesis
The story in at most 45 words.
claims
6 to 10 factual claims. Each has the claim (30 words max), the report it comes from, and a passage of at most 25 words copied exactly from that report, or empty if it cannot be quoted.
unclear
What the reports leave open or contradict each other on.
stakes
Why it matters, in at most 60 words.
watch
2 to 4 things to watch next.

The cited reports are appended to the prompt in full; they are not republished here.

Sources

  1. 1CNBCGovernment websites go dark as software flaw sparks security scramble
  2. 2France 24Government and corporate websites go dark as AI exposes critical security flaws - France 24
  3. 3Al JazeeraWorld leaders warn of security flaws found in PHP programming language | Cybersecurity News

Related: Satcove Disagreement Report 2026 · What an agreement score means · More Technology analyses