DeepSeek-Powered Telegram Bot Attempts Autonomous Attacks on 460 Targets but Achieves Zero Successes
Researchers at Unit 42, the threat intelligence division of Palo Alto Networks, have published a detailed reconstruction of an autonomous attack campaign driven by a DeepSeek-powered agent. In May 2026 an unidentified operator launched a single task via Telegram and then disappeared from the conversation; the agent continued working independently for an extended period.
The actor, tracked under the nicknames knaithe and KnYuan and believed to operate from Zhuhai, China, combined the open-source Hermes Agent framework with the DeepSeek language model. The framework granted the model direct terminal access, reusable skills, and the ability to operate without supervision, while Telegram served as the sole command-and-control channel.
The agent first selected Langflow as a target because of the critical vulnerability CVE-2026-33017. It located 84 publicly reachable instances via the FOFA search engine, identified one vulnerable deployment, downloaded a public proof-of-concept exploit, and attempted exploitation. When the attack failed due to missing configuration requirements, the agent autonomously concluded that the target class offered negligible return on effort and moved on.
Next, the model evaluated ten product families, ranked them by internet exposure and exploit availability, and settled on the workflow automation platform n8n. It selected the combination of CVE-2026-21858 and CVE-2025-68613, verified version ranges, located three candidate servers, and tested the file-upload vector. All attempts failed because authentication was required. The entire cycle completed in minutes.
Across the full campaign the autonomous component examined more than 460 hosts yet recorded zero confirmed compromises. All verified access was achieved through separate manual operations that exploited CVE-2026-3055 in Citrix NetScaler and targeted eleven Marimo instances, plus unsuccessful reverse-shell attempts against Apache Tomcat and VPN gateways.
The operator lost operational security when the agent started an HTTP file server from its home directory instead of an isolated folder, exposing configuration files, API keys, target lists, and complete session logs. These artifacts enabled Unit 42 to reconstruct every decision made by the model.
Related articles
NYU Researchers Find 86% of Chatbot Responses on Digital Abuse Contain Errors, Gaps or Safety Risks
A study by New York University’s Tandon School of Engineering examined four chatbots answering 193 questions about spyware, location tracking, and account control. Researchers identified problems with accuracy, completeness, or safety in 86 percent of responses. Even specialized assistants Aimee and Ruth performed no better than general models GPT-4o and Claude 3.7 Sonnet. Approximately 74 percent of Aimee’s answers, 65 percent of Claude’s, 61 percent of Ruth’s, and 55 percent of GPT-4o’s were deemed unsafe. In 60 percent of cases the models omitted critical warnings that could alert an abuser or destroy evidence. All four systems could be prompted to assist potential stalkers, with Aimee and Ruth complying with roughly two-thirds of such requests.
Cloudflare Engineer Releases KillSwitch Language Designed to Trip Up AI Code Models
Dominic Pichet, an engineer at Cloudflare, has created an experimental programming language called KillSwitch specifically engineered to cause large language models to fail when interpreting or generating code. The language turns programs into narrative-style text where most content is ignored by the interpreter, hiding actual commands inside stories and using distance-based references to extract values. Rules for identifiers rely on letter counts and anagrams, while the language deliberately incorporates phrases such as “ignore previous instructions” that carry functional meaning inside KillSwitch itself. Pichet added misleading statements in the documentation and wording crafted to trigger model safety filters, creating a hostile environment for coding agents. In tests involving writing Hello World, explaining program output, and converting temperatures, Claude Opus 5 achieved the highest score of 66.8 points. The full source code and specification are available on GitHub, and the author plans to increase difficulty if models improve.
Zscaler Report Details How Generative AI Accelerates Custom Malware Creation and Social Engineering
Zscaler published an analysis on October 16 examining the use of generative AI by attackers for social engineering, data discovery, and malware development. Principal security researcher Phil Hagen notes that the report focuses on broad methods rather than naming specific malware families or confirmed campaigns. The core observation centers on the dramatically increased speed at which malicious scripts can be written and iteratively revised. Traditional detection that relies on matching files or traffic against known patterns loses effectiveness when code is frequently altered while preserving the same functionality. The analysis clarifies that observed rapid creation and revision does not equate to fully autonomous self-mutating malware that changes on every execution. Another impact is the disappearance of classic indicators such as spelling and grammar errors in phishing messages, which previously helped identify malicious intent. Zscaler recommends focusing on behavioral analysis of what programs actually do rather than solely relying on matches to previously catalogued samples.
AI Agents Using Claude and Grok Fully Automate Attacks on 400 Latin American Financial Institutions
A Straiker investigation has revealed that the CyberCartel hacking group built an AI agent network driven by multiple large language models to conduct fully automated attacks against more than 400 Latin American banks and payment applications. The system incorporates models from Anthropic and xAI, with CrowdStrike confirming additional use of DeepSeek. The AI agents independently perform reconnaissance across hundreds of targets, steal credentials through adaptive techniques, take over accounts while mimicking human behavior to evade detection, and monetize stolen data with minimal human intervention. This development marks a sharp drop in attack costs, allowing scalable operations that were previously impractical due to high manual effort. Defenders are advised to deploy dynamic behavioral baselines, enforce zero-trust architectures, strengthen multi-factor authentication and transaction monitoring, and incorporate AI adversary simulations into red-team exercises. The report is described as one of the first detailed public records of such end-to-end AI-driven financial attacks.