# Google Project Zero Shortens Vulnerability Disclosure Windows Toward 30 Days Due to AI-Assisted Exploitation

> Project Zero ended its 30-day post-patch grace period on November 13, 2026 and now publishes vulnerability details 90 days after reporting, whether or not a fix exists. The window shrinks to 60 days on February 2, 2027 and 30 days on January 3, 2028, citing LLM-accelerated vulnerability discovery and exploitation.

- **Published:** 2026-11-16T16:03:25Z
- **Last reviewed:** 2026-11-16T16:03:25Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-6156
- **Card image:** https://intel.threadlinqs.com/og/threat/TL-2026-6156.png
- **ID:** TL-2026-6156
- **Severity:** INFO
- **Category:** THREAT_INTEL
- **Status:** MONITORING
- **Detections:** 9 (1 shown free below) · **IOCs:** 9

## Description

On November 13, 2026, Google Project Zero (Natalie Silvanovich) announced a new 'Full X' disclosure policy. Since 2021 the policy had been '90 + 30': up to 90 days for the vendor to remediate plus a 30-day soak period for patch rollout. Under the new policy, details are disclosed X days after the report reaches the vendor, regardless of whether a fix has shipped. X is 90 now, 60 from February 2, 2027, and 30 from January 3, 2028. Most special cases, including the grace period and the variant policy, are removed. Severity is no longer a disclosure decision factor, and Project Zero will keep tracking fix status but will not treat it as a decision point.

Project Zero's stated rationale is AI-driven. LLMs speed up vulnerability discovery and exploitation and raise the likelihood that several parties, including malicious ones, find the same bug. Attackers can reverse engineer patches quickly, so soak time gives a false sense of secrecy. Project Zero now reports several vulnerabilities a week instead of a few a year, and negotiating fix status and grace periods has become unmanageable. The policy aims to encourage faster remediation, reduce patch gap, and allow automation. SecurityWeek adds that the phased transition addresses users' exposure to unpatched flaws, vendors rushing new fixes while overlooking older, more severe bugs, and vendors that cannot meet the new timelines. Silvanovich described the goal as a schedule that is 'challenging but realistic'.

Exceptions: vulnerabilities exploited in the wild follow Google's vulnerability disclosure policy, which sets disclosure after 7 days. If details are already public from others, Project Zero may publish its own. Details are published earlier if the vendor requests it. Hardware exceptions are rare (Spectre and Meltdown); the policy page says exceptions have been granted five times in 12 years. A related Project Zero post (October 6, 2026, 'How to fix a bug in a fix') urges vendors to build emergency patching capabilities (feature flags, filtering, alternate delivery channels, hotpatching) and states that LLMs let a wider range of actors carry out novel attacks more quickly, so vendors should plan for active exploitation.

This is a policy change, not a vulnerability, malware family or actor campaign. There is no CVE, no PoC and no attributed actor. Defender relevance: the time between a vendor patch and public technical detail is shrinking, and unpatched bug details will become public 90 days after report (later 60 and 30), so patch-management and compensating-control processes must be able to react faster. IOCs listed are entities and sources only; there are no network or file indicators.

## MITRE ATT&CK

- T1587 Develop Capabilities
- T1588 Obtain Capabilities
- T1190 Exploit Public-Facing Application
- T1203 Exploitation for Client Execution

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**Web Server Worker Spawning Shell, Downloader or Recon Tool — Exploitation of Public-Facing Apps (T1190)** (`TL-2026-6156-SPL-001`)

Behaviour-first detection for TL-2026-6156 (Google Project Zero Full X disclosure policy, 90/60/30-day windows). As unpatched bug details become public sooner and LLM-assisted patch diffing shortens n-day weaponisation, exploitation of internet-facing apps (T1190) is the most likely consequence. Fires when an IIS/Apache/nginx/PHP/Tomcat worker spawns cmd (T1059.003), PowerShell (T1059.001), a download utility such as certutil, bitsadmin or curl (T1105), scripting hosts or recon tools, and labels each hit with the matching technique.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1190, T1059.001, T1059.003, T1105
- **Data sources:** sysmon, windows_security
- **Severity:** high
- **Confidence:** high

### Splunk SPL

````spl
index=windows sourcetype="XmlWinEventLog:Microsoft-Windows-Sysmon/Operational" EventCode=1 earliest=-24h
(ParentImage="*\\w3wp.exe" OR ParentImage="*\\httpd.exe" OR ParentImage="*\\nginx.exe" OR ParentImage="*\\php-cgi.exe" OR ParentImage="*\\tomcat*.exe")
(Image="*\\cmd.exe" OR Image="*\\powershell.exe" OR Image="*\\pwsh.exe" OR Image="*\\wscript.exe" OR Image="*\\cscript.exe" OR Image="*\\mshta.exe" OR Image="*\\certutil.exe" OR Image="*\\bitsadmin.exe" OR Image="*\\curl.exe" OR Image="*\\whoami.exe" OR Image="*\\net.exe" OR Image="*\\net1.exe" OR Image="*\\nltest.exe" OR Image="*\\systeminfo.exe")
```A web server worker spawning an interpreter, downloader or recon tool is the classic post-exploit signal for T1190 on unpatched public-facing apps```
| eval behaviour=case(match(Image,"(?i)powershell|pwsh"),"T1059.001_powershell", match(Image,"(?i)cmd\.exe$"),"T1059.003_cmd", match(Image,"(?i)certutil|bitsadmin|curl"),"T1105_ingress_tool_transfer", match(Image,"(?i)wscript|cscript|mshta"),"T1059_script_host", true(),"T1190_post_exploit_recon")
| stats count min(_time) as first_seen max(_time) as last_seen values(behaviour) as behaviours values(Image) as child_images values(CommandLine) as command_lines by host ParentImage ParentCommandLine User
| convert ctime(first_seen) ctime(last_seen)
| table first_seen last_seen host User ParentImage behaviours child_images command_lines count
````

### Microsoft KQL

```kql
DeviceProcessEvents
| where Timestamp > ago(24h)
| where InitiatingProcessFileName in~ ("w3wp.exe","httpd.exe","nginx.exe","php-cgi.exe","tomcat9.exe","tomcat10.exe")
| where FileName in~ ("cmd.exe","powershell.exe","pwsh.exe","wscript.exe","cscript.exe","mshta.exe","certutil.exe","bitsadmin.exe","curl.exe","whoami.exe","net.exe","net1.exe","nltest.exe","systeminfo.exe")
| extend Behaviour = case(
    FileName in~ ("powershell.exe","pwsh.exe"), "T1059.001_powershell",
    FileName =~ "cmd.exe", "T1059.003_cmd",
    FileName in~ ("certutil.exe","bitsadmin.exe","curl.exe"), "T1105_ingress_tool_transfer",
    FileName in~ ("wscript.exe","cscript.exe","mshta.exe"), "T1059_script_host",
    "T1190_post_exploit_recon")
| project Timestamp, DeviceName, AccountName, InitiatingProcessFileName, InitiatingProcessCommandLine, FileName, ProcessCommandLine, Behaviour, SHA256
| order by Timestamp desc
```

### Sigma

```yaml
title: Web Server Worker Process Spawning Shell, Downloader or Recon Tool (TL-2026-6156 Exploitation)
id: 6f1c2a54-3b7e-4d0a-9a51-1c8e5b7d2f01
status: experimental
description: Detects IIS, Apache, nginx, PHP or Tomcat worker processes spawning command interpreters, download utilities or recon utilities, the typical outcome of exploiting an unpatched public-facing application (T1190). Relevant as the Project Zero Full X disclosure policy shortens the window between bug details and weaponised n-day exploits.
references:
    - https://projectzero.google/2026/11/disclosure-policy.html
    - https://projectzero.google/vulnerability-disclosure-policy.html
    - https://www.securityweek.com/google-project-zero-sets-course-for-30-day-disclosure-deadline/
author: Threadlinqs Team
date: 2026-11-16
tags:
    - attack.initial-access
    - attack.t1190
    - attack.execution
    - attack.t1059.001
    - attack.t1059.003
    - attack.command-and-control
    - attack.t1105
logsource:
    category: process_creation
    product: windows
detection:
    selection_parent:
        ParentImage|endswith:
            - '\w3wp.exe'
            - '\httpd.exe'
            - '\nginx.exe'
            - '\php-cgi.exe'
            - '\tomcat9.exe'
            - '\tomcat10.exe'
    selection_child:
        Image|endswith:
            - '\cmd.exe'
            - '\powershell.exe'
            - '\pwsh.exe'
            - '\wscript.exe'
            - '\cscript.exe'
            - '\mshta.exe'
            - '\certutil.exe'
            - '\bitsadmin.exe'
            - '\curl.exe'
            - '\whoami.exe'
            - '\net.exe'
            - '\net1.exe'
            - '\nltest.exe'
            - '\systeminfo.exe'
    condition: selection_parent and selection_child
falsepositives:
    - Web applications that legitimately shell out for maintenance or deployment scripts
    - Administrators running diagnostics through a web-based management console
level: high

```

**Known false positives:**

- Web apps that legitimately shell out for deployment or maintenance tasks
- Admin diagnostics run through web-based management consoles

**Tuning notes:** Exclude known-good ParentCommandLine / app-pool names and scripted deployment accounts. Scope to hosts tagged as internet-facing. Add further app server parents (e.g. java.exe for specific services) if used in your estate. Prioritise the T1105 behaviour label, which usually means a payload is being staged.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- Client Application Spawning Script Host or LOLBin — Client Exploitation (T1203) (locked: unlock with TL Creds)
- Repeated Memory-Corruption Crashes of Exposed Applications (T1190, T1203, T1210) (locked: unlock with TL Creds)
- Patch Diffing / Exploit Development Tooling on Non-Research Hosts (T1587.004, T1588.005) (locked: unlock with TL Creds)
- Linux Web Server Spawning Shell or Download Utility — Web Shell / Exploitation (T1190, T1505.003) (locked: unlock with TL Creds)
- Kernel Driver Loaded from User-Writable Path or Known-Vulnerable Driver (T1068, T1562.001, T1543.003) (locked: unlock with TL Creds)
- Non-Browser Access to [indicator] Disclosure Source (T1588.005, T1596) (locked: unlock with TL Creds)
- SYSTEM Shell from Non-Service Parent or Potato-Class Tool — Local Privilege Escalation (T1068, T1134.001) (locked: unlock with TL Creds)
- Network-Facing Service Spawning Shell or Proxy-Execution LOLBin — Exploitation of Remote Services (T1210) (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sources

- [Google Project Zero Sets Course for 30-Day Disclosure Deadline (SecurityWeek, Ionut Arghire)](https://www.securityweek.com/google-project-zero-sets-course-for-30-day-disclosure-deadline/)
- [Project Zero disclosure policy announcement (Natalie Silvanovich)](https://projectzero.google/2026/11/disclosure-policy.html)
- [Project Zero vulnerability disclosure policy](https://projectzero.google/vulnerability-disclosure-policy.html)
- [How to fix a bug in a fix (Project Zero emergency patching)](https://projectzero.google/2026/10/emergency-patching.html)
- [Project Zero vulnerability disclosure FAQ (exceptions)](https://projectzero.google/vulnerability-disclosure-faq.html#exceptions)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-6156
