# Linux Foundation Akrites Initiative: Coordinated Vulnerability Disclosure Platform for AI-Enabled Open-Source Threats Reaches Operational Milestone

> The Linux Foundation's Akrites initiative — a coalition of 20+ technology, AI, and financial organizations defending critical open-source software against AI-enabled cyber threats — announced its vulnerability disclosure and remediation platform will go live in September 2026. Built on Carnegie Mellon University CERT/CC's VINCE platform with LLM-powered deduplication and patch creation capabilities, the platform will serve as a shared Security Incident Response Team (SIRT) and standardized Coordinated Vulnerability Disclosure (CVD) process for critical open-source projects. The initiative addresses the accelerating crisis where AI models can surface vulnerabilities in minutes but fewer than 5% of validated OSS vulnerabilities are being patched, and the median time to exploitation has collapsed from one year to one day.

- **Published:** 2026-08-19T00:00:00Z
- **Last reviewed:** 2026-08-19T00:00:00Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-2073
- **Card image:** https://intel.threadlinqs.com/og/threat/threat-intel-akrites-open-source-cvd-platform-tl-2026-2073.webp?v=38ff47e6 (Threadlinqs threat intel: Akrites CVD Platform for Open-Source: Analysis &… (TL-2026-2073) — Informational severity, 12 IOCs, 9 detection rules.)
- **ID:** TL-2026-2073
- **Severity:** INFORMATIONAL
- **Category:** THREAT_INTEL
- **Status:** PENDING
- **Detections:** 9 (1 shown free below) · **IOCs:** 12 (3 shown free below)

## Description

Akrites, launched on June 25, 2026 by the Linux Foundation and the Open Source Security Foundation (OpenSSF), represents a structural response to the transformation of the vulnerability landscape by AI. The initiative's name derives from the Akritai — the Byzantine Empire's frontier guardians who stood watch where defenses were thinnest. In software, that frontier is upstream open source.

The core problem Akrites addresses is a fundamental asymmetry: frontier AI models (Anthropic Claude, OpenAI GPT, Google DeepMind) can now scan a major open-source project and surface vulnerabilities in minutes, compressing what once took security experts weeks into automated machine-speed discovery. AI security tools enable five different reporters to describe the same vulnerability in a popular library within a single week. The result is a flood of duplicate, low-quality, and AI-generated reports that overwhelm volunteer maintainers — the signal-to-noise ratio collapses, real vulnerabilities get buried, and the patch window shrinks.

Anthropic's Project Glasswing (April 2026) demonstrated the scale: 23,019 vulnerabilities found across 1,000+ open-source projects in its first month, with 3,900 high/critical severity. Of 530 reported to maintainers, only 75 were patched. Endor Labs, a founding member, reported that fewer than 5% of validated OSS vulnerabilities surfaced in recent months have been patched. J.P. Morgan's 'Patchmageddon' report documented that median time to exploitation fell from approximately one year in 2021 to one day in 2026, with 80% of exploitations occurring on or before disclosure day. Tuskira Research found vulnerability discovery outpaced patching by 16.5x.

Akrites operates as a single, confidential front door for vulnerability reports against critical open-source projects. The platform builds on Carnegie Mellon University CERT/CC's VINCE (Vulnerability Information and Coordination Environment) — a Django-based web application launched in 2020 that replaced CERT/CC's legacy PGP-encrypted email coordination model. Akrites augments VINCE with LLM capabilities for deduplication (an estimated 30% of incoming reports are duplicates) and automated patch creation assistance.

The vulnerability flow follows a four-stage process: (1) Intake — a finding surfaces to the SIRT, classified TLP:RED from the start, visible only to the case team; (2) Deduplicate and Validate — the SIRT merges duplicates, validates severity, and assigns ownership; (3) Remediate — maintainers and coalition engineers prepare and test the fix, held as TLP:RED case material; (4) Synchronized Disclosure — the upstream project enters one CVD window, and the fix publishes to the original namespace at disclosure.

Founding signatories span 20+ organizations across three tiers: AI frontier labs (Anthropic, OpenAI); cloud and tech giants (Amazon Web Services, Google, Microsoft/GitHub, IBM, NVIDIA, Cisco, Ericsson); cybersecurity firms (Chainguard, Endor Labs, RapidFort, Sonatype, Zscaler); and large enterprises (Citi, JPMorganChase, Vodafone, Red Hat, Rust Foundation). Members commit 1-10 engineers and pay tier-based fees (Premier at $250,000, General at $200,000, or Associate at $0 for recognized open-source foundations).

Akrites also serves as a 'maintainer of last resort' for abandoned-but-critical packages — where a critical package has no active maintainer, Akrites will step in to ensure fixes reach the latest version. Seed funding comes from Alpha-Omega, the Linux Foundation's directed fund that has awarded over $20 million in 70+ grants since its inception, including $5.8 million across 14 projects in 2025 alone.

Key principles include: confidentiality-first (based on TLP 2.0 protocol); technical contribution as the price of admission; synchronized disclosure (no participant gets a head start); least privilege across the program; and one maintainer-facing front door with fixes published into the package's original namespace. The initiative operates in hardened infrastructure with isolated secure enclaves for vulnerability analysis, secure VM analyst workbenches, MFA, and monitoring.

The platform's go-live in September 2026 follows a summer of escalating community debate. Filippo Valsorda, former Go Security lead, published 'Vulnerability Reports Are Not Special Anymore' on June 23, 2026, arguing that LLMs have collapsed the discovery bottleneck and that coordinated disclosure's central premise — that disclosure buys defenders time — is increasingly false. The curl project ended its bug bounty program in January 2026 (citing 'AI slop reports' overwhelming maintainers, with confirmed vulnerability rates dropping below 5%) and suspended vulnerability reporting entirely for July 2026. On the oss-security mailing list in August 2026, community members raised concerns about Akrites potentially fragmenting existing coordination processes like the distros list, while others clarified that Akrites operates across all four stages of vulnerability handling while the distros list only handles synchronized disclosure.

Akrites positions itself as complementary to vulnerability-finding efforts (Project Glasswing, MITRE/CVE, Lightwell, FIRST) rather than competing with them — it focuses on coordinating the disclosure of findings and can accept reports from any of these programs. The platform will be open-sourced once finalized, available for anyone to use for their own purposes. The initiative's tagline is 'Patch the commons, together.'

## MITRE ATT&CK

- T1595.002 Active Scanning: Vulnerability Scanning
- T1588.002 Obtain Capabilities: Tool
- T1588.006 Obtain Capabilities: Vulnerabilities
- T1583.001 Acquire Infrastructure: Domains
- T1583.003 Acquire Infrastructure: Virtual Private Server
- T1583.006 Acquire Infrastructure: Web Services
- T1190 Exploit Public-Facing Application
- T1071.001 Application Layer Protocol: Web Protocols
- T1499.004 Endpoint Denial of Service: Application or System Exploitation

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

**OSS Vulnerability Rapid Exploitation — Post-Disclosure Zero-Day Window Abuse** (`TL-2026-2073-SPL-002`)

Detects exploitation attempts targeting known open-source software vulnerabilities within the critical post-disclosure window. The Akrites research documents that median time-to-exploitation has collapsed from ~1 year (2021) to ~1 day (2026), with 80% of exploitations occurring on or before disclosure day. This detection identifies exploit attempts against specific OSS components (web frameworks, CMS platforms, CI/CD tools, application servers) within the window where Akrites-coordinated patches are being prepared but not yet deployed. It monitors for known exploitation patterns — deserialization, path traversal, remote code execution, SQL injection — against the specific vulnerable endpoints documented in recent CVE disclosures for OSS projects.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1190, T1499.004
- **Data sources:** web_server_logs, waf_logs, ids_ips, proxy_logs
- **Severity:** critical
- **Confidence:** high

### Splunk SPL

```spl
index=* sourcetype=*access_combined earliest=-24h
  (
    /* Known OSS exploitation patterns — Log4j, Struts, Spring4Shell, etc. */
    uri_path IN ("*/jndi*", "*/InvokeAction*", "*/spring4shell*", "*/class.module.classLoader*", "*/catalina/exec*", "*/jboss-seam-2*", "*/cgi-bin*", "*/wp-content/plugins/*", "*/wp-content/themes/*", "*/modules/*", "*/includes/*")
    OR uri_query IN ("*class.module.classLoader*", "*jndi:ldap*", "*jndi:rmi*", "*jndi:dns*", "*${*", "*#{*", "*Runtime.getRuntime*", "*ProcessBuilder*", "*java.lang.Runtime*")
  )
  status IN (200, 302, 400, 403, 500)
| eval exploit_category = case(
    like(uri_path, "%/jndi%") OR like(uri_query, "%jndi:%") OR like(uri_query, "%${%") OR like(uri_query, "%#{%"), "jndi_injection_log4j",
    like(uri_path, "%/InvokeAction%"), "struts2_rce",
    like(uri_path, "%class.module.classLoader%") OR like(uri_path, "%spring4shell%"), "spring4shell_rce",
    like(uri_path, "%catalina/exec%") OR like(uri_path, "%jboss-seam-2%"), "tomcat_jboss_exploit",
    like(uri_query, "%Runtime.getRuntime%") OR like(uri_query, "%ProcessBuilder%") OR like(uri_query, "%java.lang.Runtime%"), "deserialization_rce",
    1=1, "other_oss_exploit"
  )
| stats count as exploit_attempts dc(src_ip) as unique_sources values(src_ip) as attackers values(status) as http_statuses earliest(_time) as first_seen latest(_time) as last_seen by exploit_category, uri_path
| where exploit_attempts > 1
| eval threat_window = tostring(round((last_seen - first_seen) / 60, 1)) + " minutes"
| eval severity = case(
    exploit_attempts > 100, "CRITICAL",
    exploit_attempts > 20, "HIGH",
    exploit_attempts > 5, "MEDIUM",
    1=1, "LOW"
  )
| lookup cve_info.csv OUTPUT cve_id, cvss_score, known_ransomware_use
| eval context = "OSS vulnerability exploitation within collapsed disclosure window. 80% of exploitation occurs on or before disclosure day (Patchmageddon). Verify patch status for affected OSS components."
| table exploit_category, uri_path, exploit_attempts, unique_sources, attackers, threat_window, severity, context
| sort - exploit_attempts
```

### Microsoft KQL

```kql
let exploit_window = 24h;
let exploit_patterns = dynamic([
    "jndi:ldap", "jndi:rmi", "jndi:dns", "jndi:iiop",
    "${jndi", "${env:", "${sys:",
    "#{${java",
    "class.module.classLoader",
    "spring4shell",
    "InvokeAction",
    "Runtime.getRuntime",
    "ProcessBuilder",
    "java.lang.Runtime",
    "catalina/exec",
    "jboss-seam-2",
    "../../../../../etc/passwd",
    "..%252f..%252f..%252f",
    "__proto__",
    "constructor[prototype]"
]);
let known_webshell_paths = dynamic([
    "/modules/", "/includes/", "/wp-content/plugins/", "/wp-content/themes/",
    "/vendor/", "/node_modules/", "/uploads/"
]);
union AWSCloudTrail, W3CIISLog, NGXAccessLog
| where TimeGenerated >= ago(exploit_window)
| where CSUriQuery has_any (exploit_patterns) or CSUriStem has_any (known_webshell_paths)
| extend ExploitCategory = iif(
    CSUriQuery has "jndi:" or CSUriQuery has "${jndi" or CSUriQuery has "${env:" or CSUriQuery has "${sys:", "JNDI/Log4j Injection",
    CSUriQuery has "class.module.classLoader" or CSUriQuery has "spring4shell", "Spring4Shell RCE",
    CSUriQuery has "InvokeAction", "Struts2 RCE",
    CSUriQuery has "Runtime.getRuntime" or CSUriQuery has "ProcessBuilder" or CSUriQuery has "java.lang.Runtime", "Deserialization RCE",
    CSUriStem has "../../", "Path Traversal",
    CSUriQuery has "__proto__" or CSUriQuery has "constructor[prototype]", "Prototype Pollution",
    "Other OSS Exploit"
  )
| summarize Attempts = count(), DistinctIPs = dcount(CIP), FirstSeen = min(TimeGenerated), LastSeen = max(TimeGenerated) by ExploitCategory, CSUriStem
| where Attempts > 1
| extend ExposureWindowMinutes = datetime_diff('minute', LastSeen, FirstSeen)
| extend ImpactLevel = case(
    Attempts > 100, "Critical",
    Attempts > 20, "High",
    Attempts > 5, "Medium",
    "Low"
  )
| extend Context = "Exploitation of OSS vulnerability within collapsed disclosure window. J.P. Morgan Patchmageddon: 80% of exploitations occur on/before disclosure day. Verify patch status."
| project ExploitCategory, CSUriStem, Attempts, DistinctIPs, ExposureWindowMinutes, ImpactLevel, Context
| order by Attempts desc
```

### Sigma

```yaml
title: OSS Vulnerability Rapid Exploitation — Post-Disclosure Zero-Day Window
id: b7c8d9e0-1f2a-3b4c-5d6e-7f8a9b0c1d2e
status: experimental
description: Detects exploitation attempts against known open-source software vulnerabilities within the critical post-disclosure window. The Akrites research context documents that median time-to-exploitation has collapsed to approximately one day, with 80% of exploitations occurring on or before disclosure day. Monitors for known exploitation patterns — JNDI injection, deserialization, path traversal, RCE — targeting OSS components.
references:
  - https://www.jpmorgan.com/insights/cybersecurity/patchmageddon
  - https://www.anthropic.com/news/project-glasswing
  - https://attack.mitre.org/techniques/T1190/
author: AII-Detector — ThreadLinqs Intelligence
date: 2026-08-19
tags:
  - attack.initial-access
  - attack.t1190
  - attack.t1499.004
  - cve.2026
  - exploit.kit
  - web-exploitation
  - zero-day
logsource:
  category: web
  product: proxy
  service: http
detection:
  selection_jndi:
    uri_query|contains:
      - 'jndi:ldap://'
      - 'jndi:rmi://'
      - 'jndi:dns://'
      - '${jndi:'
      - '${env:'
      - '#{${java:'
  selection_rce:
    uri_query|contains:
      - 'Runtime.getRuntime'
      - 'ProcessBuilder'
      - 'java.lang.Runtime'
      - 'class.module.classLoader'
  selection_path_traversal:
    uri_path|contains:
      - '../../../../../etc/passwd'
      - '..%252f..%252f..%252f..%252f'
  timeframe: 1h
  condition: 1 of selection_*
  aggregation:
    field: src_ip
    type: count
    value: 3
    condition: gt
falsepositives:
  - Security research and bug bounty testing
  - Automated vulnerability scanners in authorized pen-testing
  - Honeypot interactions and threat intelligence platform probes
  - WAF/IPS false positives during legitimate traffic
level: high
```

**Known false positives:**

- Authorized penetration testing and red team exercises
- Vulnerability scanners in authorized security assessments
- Automated security research bots scanning for known vulnerabilities
- WAF/IPS rules triggering on decoy or test payloads

**Tuning notes:** Establish baseline exploit attempt counts per OSS component for your environment. Create allowlists for authorized security testing sources. Correlate with patch management system to prioritize detections for unpatched OSS components. Integrate with CISA KEV feed to dynamically escalate detections matching known exploited vulnerabilities. Adjust time window downward if detection volume is too high.

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

- AI-Driven Vulnerability Scanning — Suspicious User-Agent and Request Pattern Detection (locked: unlock with TL Creds)
- Post-Exploitation via Unpatched OSS Components — Suspicious Process and File System Indicators (locked: unlock with TL Creds)
- OSS Vulnerability-Driven Lateral Movement — Suspicious Authentication and Network Propagation (locked: unlock with TL Creds)
- OSS Vulnerability Scanning — AI-Scale Automated Reconnaissance (locked: unlock with TL Creds)
- Open Source Vulnerability Exploitation — Webshell Deployment and Backdoor Persistence (locked: unlock with TL Creds)
- OSS C2 Infrastructure — Domain and VPS Acquisition for Web-Protocol Command and Control (locked: unlock with TL Creds)
- AI-Scale Automated Vulnerability Scanning — OSS Application Reconnaissance Detection (locked: unlock with TL Creds)
- C2 Beaconing via Web Protocols — Compromised OSS-Dependent Systems (locked: unlock with TL Creds)

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

## Sample indicators (3 of 12)

Defanged, TLP:CLEAR.

- Domain: `akrites[.]org`
- Domain: `kb[.]cert[.]org`
- Domain: `words[.]filippo[.]io`

Full set of 12 indicators unlocks with TL Creds: https://threadlinqs.com/tl-creds

## Sources

- [Linux Foundation and Industry Leaders Launch Akrites](https://www.linuxfoundation.org/press/linux-foundation-and-industry-leaders-launch-akrites-to-defend-critical-open-source-software-against-ai-enabled-cyber-threats)
- [Exclusive: Linux Foundation's Akrites to Go Live in September](https://www.infosecurity-magazine.com/news/linux-foundations-akrites-go-live/)
- [Akrites Official Website](https://akrites.org/)
- [Akrites: Centralizing Open-Source Vulnerability Disclosure (Groundy Analysis)](https://groundy.com/articles/linux-foundation-akrites-centralizes-open-source-vulnerability-disclosure/)
- [Vulnerability Reports Are Not Special Anymore (Filippo Valsorda)](https://words.filippo.io/vuln-reports/)
- [The End of the Curl Bug-Bounty (Daniel Stenberg)](https://daniel.haxx.se/blog/2026/01/26/the-end-of-the-curl-bug-bounty/)
- [Curl Summer of Bliss (Daniel Stenberg)](https://daniel.haxx.se/blog/2026/06/15/curl-summer-of-bliss/)
- [Linux Foundation Announces $12.5M in Grant Funding for Open Source Security](https://www.linuxfoundation.org/press/linux-foundation-announces-12.5-million-in-grant-funding-from-leading-organizations-to-advance-open-source-security)
- [Anthropic Launches Project Glasswing](https://www.anthropic.com/news/project-glasswing)
- [CERT/CC Releases VINCE Software Vulnerability Collaboration Platform](https://www.sei.cmu.edu/news/certcc-releases-vince-software-vulnerability-collaboration-platform/)
- [oss-security Mailing List: Akrites Discussion (August 2026)](https://www.openwall.com/lists/oss-security/2026/08/04/32)
- [J.P. Morgan Patchmageddon Report (Eye on the Market, 2026)](https://www.jpmorgan.com/insights/cybersecurity/patchmageddon)
- [Overrun with AI Slop, cURL Scraps Bug Bounties (Ars Technica)](https://arstechnica.com/security/2026/01/overrun-with-ai-slop-curl-scraps-bug-bounties-to-ensure-intact-mental-health/)
- [OSS Security Initiatives Comparison](https://tzafaar.codeberg.page/other/oss-security-initiatives-comparison.html)
- [Alpha-Omega: Scaling Open Source Security with AI](https://alpha-omega.dev/blog/scaling-open-source-security-with-ai/)

## 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-2026-2073