Correction, Oct 25: Palo Alto Networks Unit 42 has retracted the Oct 23 update this item was based on. Unit 42 now says what it saw was an attack on a European government agency through a vulnerable PHP application, and that it is unable to confirm the initial access came through the PHP2Shell CVEs (CVE-2026-109618/109619). The rest of its brief (patch now; Threat Prevention signatures) stands. Treat the details below as describing an intrusion of unconfirmed cause, not a confirmed PHP2Shell breach. PHP2Shell exploitation attempts are still reported by other sources, and patching remains urgent.

Original item (Oct 23), as published:

Palo Alto Networks Unit 42 updated its PHP2Shell (CVE-2026-109618/109619) brief on Oct 23 with the first confirmed intrusion it has published, as opposed to scanning or exploit attempts. Unit 42 says it is still not a widespread outbreak, but the bugs are now exploited with real post-exploitation activity.

• Victim: a government entity in Europe (not named).
• Timeline: within two hours of initial exploitation the attackers had read password hashes and multi-factor authentication secrets, set up an NTLM-relay tunnel into Active Directory and attempted Domain Admin escalation.
• Agentic attacker: Unit 42 says the behavior strongly indicates an AI agent driving the web shell: command batches wrapped in delimiters such as ==SHELLUSER== and ==PASSWD== so a model can parse the output, short serially spaced tasks, and quick pivots after failures.
• Exposure: Cortex Xpanse counts over 22.6 million internet-exposed instances potentially vulnerable.

If you have not patched to PHP 8.5.12, 8.4.27, 8.3.36 or 8.2.35, treat any public PHP host as possibly compromised and hunt before you patch (see Threat Hunting). Patching does not remove a web shell or reset stolen MFA seeds.