AI Is Being Used in Vulnerability Research and Active Cyber Operations

Summary
The article links AI-assisted vulnerability research and attacks, reporting PHP vulnerability scanning, JPCERT/CC findings in Japan, and Microsoft’s account of Storm-0940 using ARTEX. It stresses that scanning and failed logons alone do not prove compromise.
Key points
- CVE-2026-109618 and CVE-2026-109619 can enable unauthenticated remote code execution, including in products with embedded PHP. Darktrace observed widespread scanning but no direct evidence of exploitation.
- On October 21, Darktrace recorded a six-month high in detections consistent with zero-day exploitation activity, up 64% against historic data; it did not confirm a link to the PHP vulnerabilities.
- In a controlled simulation, researchers demonstrated that Claude’s memory stores feature could be abused after an attacker gained access to establish a C2 channel and assign tasks to internal agents. Darktrace did not claim this technique was seen in operational attacks.
- JPCERT/CC reported that attackers used AI to scan for known vulnerabilities and attack APIs in incidents affecting organizations in Japan; its infrastructure analysis found Android application analysis, API discovery, and testing for weaknesses including SSRF.
- Microsoft reported that Storm-0940 used ARTEX against hundreds of organizations, including in activity attributed to the ConnectWise supply-chain compromise. Some operations reached privileged cloud access in under three hours.
- Microsoft also documented more than 151,000 failed logons across 327 accounts at eight organizations, with no successful authentication or related access changes in that campaign. Darktrace’s 65 failed authentication attempts from infrastructure later associated with the campaign did not establish compromise.
- The article emphasizes assessing what follows initial scanning or authentication attempts, including process execution, file changes, outbound traffic, and identity or directory activity.
Article Details
- Event Type
- Security news covering AI-assisted vulnerability research and attacks, PHP vulnerability scanning, and Microsoft-attributed use of ARTEX by Storm-0940.
- Impact
- Darktrace observed widespread scanning related to the PHP vulnerabilities but reported no direct evidence of their exploitation; a separate increase in zero-day-exploitation detections was not confirmed as related. JPCERT/CC reported AI-assisted vulnerability scanning and API attacks affecting organizations in Japan. Microsoft attributed the ConnectWise supply-chain compromise to Storm-0940 using ARTEX. Microsoft reported more than 151,000 failed logons in one campaign with no successful authentication, while other operations reached privileged cloud access in less than three hours.
Indicators of compromise
| Type | Indicator | Context |
|---|---|---|
| IPV4 | 144[.]172[.]64[.]128 | Source address later identified by Microsoft as performing sign-ins and Microsoft Graph activity in ConnectWise customer environments; Darktrace recorded repeated failed authentication attempts from it. |
MITRE ATT&CK
T1046 · Network Service DiscoveryIn a controlled simulation, an external operator assigned network-discovery tasks to internal Claude Code agents.T1110.003 · Password SprayingMicrosoft reported a password-spray campaign against accounts across multiple organizations.T1595.002 · Vulnerability ScanningJPCERT/CC reported attackers used AI to scan for known vulnerabilities.
CVE
CVE-2026-109618The disclosure of CVE-2026-109618 and CVE-2026-109619 placed immediate pressure on organizations running PHP. Both vulnerabilities can allow unauthenticated remote code execution, including in products and appliancesCVE-2026-109619The disclosure of CVE-2026-109618 and CVE-2026-109619 placed immediate pressure on organizations running PHP. Both vulnerabilities can allow unauthenticated remote code execution, including in products and appliances
Threat Actors
Vendors
CloudflareCloud had analyzed more than 30 million commits across more than 30,000 codebases. According to OpenAI, Cloudflare uses models accessed through Daybreak to investigate customer-authorized code and combines findingsConnectWiseactor active since 2021, used ARTEX against hundreds of organizations, and Microsoft attributed the ConnectWise supply-chain compromise [8].DarktraceDarktrace observed widespread scanning across monitored environments involving hundreds of distinct source IP addresses that public reporting has associated with activity targeting the PHP vulnerabilities. The volume ofMicrosoftJPCERT/CC does not identify ARTEX by name in that publication. The named ARTEX attribution comes from separate investigations, including Microsoft's analysis of Storm-0940.OpenAIThe PHP disclosure arrives as AI systems have become more capable at vulnerability research [2]. OpenAI launched Daybreak in June 2026, combining GPT-5.5-Cyber, Codex Security and partnerships with security researchers
Products
ClaudeIn a controlled simulation, researchers demonstrated that, in that post-compromise scenario, the Claude platform’s memory stores feature could be abused to establish a live C2 channel between an externalClaude Codefeature could be abused to establish a live C2 channel between an external operator and internal Claude Code agents. The external operator was able to dynamically assign network discovery, credential access, andCodex Securityat vulnerability research [2]. OpenAI launched Daybreak in June 2026, combining GPT-5.5-Cyber, Codex Security and partnerships with security researchers and software maintainers. OpenAI says its models have beenCodex Security CloudOpenAI reported in August that Codex Security Cloud had analyzed more than 30 million commits across more than 30,000 codebases. According to OpenAI, Cloudflare uses models accessed through Daybreak to investigateDaybreakdisclosure arrives as AI systems have become more capable at vulnerability research [2]. OpenAI launched Daybreak in June 2026, combining GPT-5.5-Cyber, Codex Security and partnerships with security researchers andPHPThe disclosure of CVE-2026-109618 and CVE-2026-109619 placed immediate pressure on organizations running PHP. Both vulnerabilities can allow unauthenticated remote code execution, including in products and appliances
