Issue 49
AI-discovered critical PHP flaw takes government websites offline worldwide
According to the AP, an AI-discovered PHP vulnerability may affect about 70 percent of the world's websites, and governments and companies worked through the night to patch it. According to Engadget, EU and Canadian government sites were also taken offline, and according to the Insurance Journal, attack attempts numbered in the millions.
Who is affected and how
- security leader
- Must immediately assess where PHP runs across the estate and speed up patching.
- security operations
- Responsible for urgent patching and monitoring logs for attack attempts.
- developer
- Applications built on PHP must be updated and tested against the patched version.
- business leader
- Websites and services running PHP carry a risk of outages and attacks.
- public-sector decision-maker
- Some government sites were taken offline because of the flaw, so decisions are also needed on the outage and on recovery.
French supermarket fridges hit by attacks enabled by basic security flaws
France's cybersecurity agency found that a lack of basic security precautions enabled attacks on French supermarkets that destroyed food and caused fires. Reports say the attack affected around 700…
security leaderThe baseline protection of refrigeration and building management systems must be checked.security operationsThe exposure of internet-connected industrial and building-services equipment must be assessed.business leaderRetail sites can suffer physical damage and food losses.vulnerability · globalCritical FortiMail zero-day (CVE-2026-104286) exploited in the wild
According to the article, attackers are exploiting a critical zero-day flaw in FortiMail, tracked as CVE-2026-104286, in live attacks. The news was published on 22 October 2026.
security leaderOrganisations using FortiMail need an immediate risk assessment.security operationsCheck for compromise and apply protective measures until a patch is released.incident · globalLiteLLM hit by second supply chain attack this year as version 1.107.0 carried malware
According to Help Net Security, LiteLLM version 1.107.0 was compromised on PyPI and Docker and steals cloud logins, developer secrets and AI keys. It is the project's second supply chain attack this…
security leaderOrganisations using LiteLLM need to rotate all affected keys and secrets.security operationsInstallations of version 1.107.0 must be identified and access credentials rotated.developerDependencies of AI infrastructure should be pinned and verified.vulnerability · globalGitLab patches maximum-severity code execution vulnerability
GitLab has released security updates for a maximum-severity flaw affecting its regular expression compiler, which allows remote attackers to execute arbitrary code. The news was reported on 22…
security operationsVulnerable GitLab servers must be updated immediately.developerSource code repositories and CI/CD environments carry a risk of compromise.Hungary
More key stories of the week
China-linked AI agent breached 2,250 Microsoft 365 tenants within hours, bypassing MFA
According to the source article's title, a China-linked AI agent breached 2,250 Microsoft 365 tenants within hours, and MFA did not apply. The news was reported on 22 October 2026.
security leaderMFA alone did not protect in this scenario, so tenant configurations should be reviewed.security operationsMicrosoft 365 tenant logs and access rights should be reviewed.research · globalAnalyses examine indirect prompt injection and trust gaps in AI agents
Two analyses examine indirect prompt injection: one uses a lab-built example similar to publicly documented patterns, and the other says systems treat retrieved text as both data and instructions…
security leaderMust treat agents' access rights and the handling of external content as a risk.developerRetrieved content must not be processed as trusted instructions in agents.journalist and researcherThe analyses give an accessible description of attack patterns against agents.incident · globalChinese hacker uses AI agents to breach dozens, or more than 100, company websites
According to Hungarian-language reports, a Chinese hacker used AI agents and a few thousand dollars to breach dozens, or more than 100, company websites. The articles were published on 20–22 October…
security leaderAI agents significantly lower the cost and the expertise required for an attack.business leaderLarge-scale attacks against company websites can be launched on a small budget.attribúció · Európai UnióKremlin warned Germany of further attacks after Leipzig airport sabotage, WSJ reports
According to a WSJ report carried by Meduza, the Kremlin told Berlin that the sabotage attempt uncovered at Leipzig airport would not be the last attack. German authorities had previously exposed a…
security leaderTransport and critical-infrastructure operators must expect further sabotage attempts.public-sector decision-makerThe state-level threat reinforces the need for infrastructure protection and international coordination.incidentHackers used Claude and Grok to automate attacks on Latin American banks
According to Market Chatter, hackers used the Claude and Grok AI models to automate attacks on Latin American banks. The news appeared on 21 October 2026.
security leaderIn the financial sector, attackers use AI to speed up and automate operations, which tests the pace of defence as well.security operationsAutomated attacks make it worth speeding up detection and response processes.case · United StatesThird-party evaluators find vulnerabilities in OpenAI monitoring system during pilot
Third-party evaluators found vulnerabilities in OpenAI's monitoring system during a test pilot. Sam Altman has promised that independent evaluators will receive "employee-like access" at the company.
security leaderIndependent AI safety evaluation may also expose gaps in internal monitoring.journalist and researcherIndependent evaluators may gain deeper access to the AI developer.incident · globalEverest extortion group claims 590 GB of data from a PwC vendor
The Everest extortion group claims it obtained 590 GB of data in an incident involving a PwC-linked vendor. The news appeared on 23 October 2026.
security leaderBecause the vendor data is exposed, third-party access must be reviewed.business leaderA supply-chain incident can also affect customer data.vulnerability · United StatesCISA adds Collabora Online server flaw to KEV after attackers backdoor Nextcloud servers
CISA has added the Collabora Online CODE Server flaw CVE-2026-110447 to its KEV catalog because attackers used proxy.php to install backdoors on Nextcloud servers. According to the headline of one…
security leaderThe absence of a patch available for a month points to weak patch management.security operationsUnpatched Nextcloud/Collabora servers must be updated and checked for traces of the backdoor.report · globalExperts warn of cybersecurity 'vulnerability tsunami' unleashed by AI
According to a Business Insider Africa roundup, executives and AI analysts are warning of major cybersecurity threats. The experts say critical systems in energy, water and healthcare may be…
security leaderPlanning patching capacity and prioritisation becomes the key issue.public-sector decision-makerThe resilience of critical systems requires policy attention.vulnerability · globalThird Arista VCO zero-day exploited within four months
Attackers are exploiting a third zero-day vulnerability in Arista VCO within four months. The news appeared on 23 October 2026.
security operationsOrganisations using Arista VCO must expect repeated zero-day exposure.law enforcementRussian hacker arrested over attack that led to a patient's death in a hospital
According to the Hungarian-language report, a Russian hacker was arrested whose attack led to a patient's death in a hospital. The article appeared on 19 October 2026.
security leaderAttacks on healthcare can endanger lives.security operationsThe continuity of hospital systems must receive priority protection.incident · European UnionMore than 421,000 AfD emails leaked after cyberattack
According to Hungarian-language reports, more than 421,000 emails leaked from the AfD after a cyberattack. The articles appeared on 20 October 2026.
security leaderProtecting email archives is a priority for political organisations.journalist and researcherAnalysing the leaked correspondence raises ethical and source-verification questions.In brief
Weekly email
One email a week with the main stories from the Agent Age editions. You confirm by email and can unsubscribe in one click.
Almost done: we sent you an email. Open it and press the confirm link. If it does not arrive within a few minutes, check your spam folder.
Please enter a valid email address and tick the consent box.
Something went wrong. Please try again later.