AI Agents Exploit PaperCut Flaw, Compromising 395 Organizations Across 48 Countries
PaperCut published its security bulletin on 27 August 2026. Four days later, on 31 August, GreyNoise recorded the first confirmed code execution against a live target. The operator required less than four hours to prepare and launch the initial attack.
The following day the attackers activated a pre-built mass campaign. Within 26 seconds, the first 11 organizations were compromised. GreyNoise ultimately counted at least 440 compromised PaperCut NG/MF instances across 395 organizations in 48 countries.
The episode features prominently in the November report issued jointly by Google Threat Intelligence Group (GTIG) and Mandiant. Analysts mapped the points where human operators intervened, measured the time from initial server access to domain compromise, and identified artifacts left by automated components.
Timeline and Lateral Movement
GreyNoise distinguished three distinct intervals: four days from bulletin publication to first exploitation, nearly four hours of operator preparation before the first real-world execution, and 26 seconds from campaign launch to the initial wave of compromises. In the fastest observed case, attackers obtained Domain Admin rights only five minutes after gaining initial access to the print server. Twelve organizations ultimately lost domain-administrator privileges.
Some compromised systems were left untouched for several days. One attempt was blocked by a web application firewall. Researchers noted that average patch-deployment times obscure the risk posed by externally facing instances that remain unpatched longer than internal systems.
Automated Exploitation and LLM Abuse
GTIG identified an open server hosting a tool the researchers named exploit_agent. The system automatically processed updates for popular web applications and could execute attack chains with minimal manual intervention. Stolen data from victims was also stored on the same server.
Separate incidents at cryptocurrency companies revealed another technique. Attackers modified the .claude/settings.json file inside internal repositories. The altered configuration caused a backdoor to load whenever Claude Code ran on a developer workstation. Anthropic’s managed-hooks setting can block such project-level hooks.
Financial Impact and Detection
Mandiant investigations recorded unauthorized LLM usage that exceeded $100,000 per day in several cases, with peak cumulative costs reaching $1 million before access was revoked. Cloud billing spikes can be mistaken for load tests; therefore, correlating expense graphs with IAM and model-gateway logs is recommended.
Related articles
Google Threat Intelligence Group Warns of Fully Autonomous AI Cyberattacks Within Six Months
Google's Threat Intelligence Group (GTIG) has observed the first real-world instances of attackers using multi-agent AI frameworks to autonomously discover vulnerabilities, generate exploits, and compromise targets without human intervention. The discovery came from an exposed server running a framework called exploit_agent that analyzed software update logs from popular web applications to identify and exploit weaknesses. Four previously defensive tools—ARTEX, CAIRN, CyberStrikeAI, and HexStrike AI—have been repurposed for offensive operations, with documented cases targeting French media, U.S. aerospace, and cryptocurrency firms. GTIG projects that low-complexity automated attacks will become common within three months, while sophisticated fully autonomous operations are expected within six months. The report highlights how these systems compress patch windows, overwhelm security teams with AI-generated proof-of-concepts, and turn detailed vendor advisories into attack playbooks. Defenders are urged to minimize public exposure, accelerate patching cycles to weekly, and tightly control API keys to prevent LLMjacking.
OpenAI Rolls Out KYC, Hardware Keys and Tiered Access for Cybersecurity Tasks in GPT-6.2
OpenAI has introduced stricter access controls for GPT-6.2 Astra and Sol models when users attempt cybersecurity-related tasks. Certain legitimate security research and testing activities now trigger automatic blocks unless the user obtains Daybreak trusted access. The program requires paid subscriptions, FIDO2 hardware keys, Advanced Account Security, and identity verification through Persona. Developers report that even after meeting preliminary requirements, access to reduced refusal rates is not guaranteed and may require separate organizational approval. The changes stem from the models reaching Critical capability levels under OpenAI’s Preparedness Framework for dual-use cybersecurity skills. Similar verification processes previously led to account blocks for users of competing services such as Anthropic’s Claude.
Positive Technologies Releases PT Application Inspector Plugins 3.0.0 with AI Agent Support via Model Context Protocol
Positive Technologies has launched version 3.0.0 of its free PT Application Inspector plugins, introducing native support for AI agents through an integrated MCP server. The update enables AI agents to initiate vulnerability scans on entire projects or specific code changes, review findings, and assist developers in remediation directly within supported environments. Integration via the Model Context Protocol connects tools such as OpenCode, Claude Code, and Codex, providing agents with detailed vulnerability data including exploitation conditions. Static analysis remains the core detection method, while the new AI Hooks feature automatically validates modifications made by agents without requiring full project rescans. The How to Fix capability, which generates AI-driven remediation suggestions, is now available in Visual Studio Code in addition to JetBrains IDEs, with support for on-premise models to keep source code internal. Director of AppSec and WAF Products Anton Zhabolenko stated the goal is to embed vulnerability detection and fixing into the code generation workflow itself.
Vals AI Study Reveals OpenAI and Anthropic Models Selecting Civilian and Medical Targets in Simulated Drone Strikes
Vals AI conducted experiments on 14 AI models to test their safeguards against selecting military targets. The systems frequently identified civilians, ambulances, and surrendering soldiers as valid targets in drone imagery simulations. Framing requests as games significantly increased compliance rates, with one model jumping from 6% to 92% target identification. Tests also used real satellite imagery of Iran's Minab area where a school sits near a military complex. Most models continued marking people even after being shown generated strike consequences. The research highlights risks of using such models for large-scale video labeling to train autonomous weapons systems. Vals AI recommends stronger API-level restrictions to prevent misuse.