Parameter Drift in n8n Workflows Allows Approved Action A to Trigger Unrelated Action B in Bitrix24
An engineer writing for Habr examined a practical security gap in AI agent orchestration using n8n, Groq, MCP, and Bitrix24. The core finding was that a human could approve one specific action while the downstream execution node attempted a different action because the two nodes sourced their parameters independently.
The test involved placing a human approval step before any write operation that would change a synthetic task title in Bitrix24. In early workflow versions the approval screen displayed one set of values while the MCP client node that performed the update_task call used its own configuration. This produced the observed drift: APPROVED A ≠ ATTEMPTED B. Bitrix24 rejected the mismatched request, so the task state remained unchanged, yet the architectural defect was confirmed.
Evidence collected during testing
The researcher recorded concrete observations across multiple runs. In one execution the human approved a title change labeled Action A, yet the downstream node attempted Action B. Bitrix24 returned a rejection and no state change occurred. The experiment deliberately varied approval configuration, execution path, fresh-state comparison, and data extraction logic to isolate the drift condition from other failure modes.
Importantly, the AI agent itself operated under read-only Bitrix24 tools. It could propose changes but could not execute writes. Authority for the actual update resided in a separate deterministic MCP client node that ran only after the approval gate. The drift therefore originated in the orchestration wiring rather than in any autonomous decision by the language model.
Action Envelope and verification steps
The corrected design introduced a single Action Envelope object containing target_system, task_id, operation, expected_title, and requested_title. Both the approval message and the subsequent update_task call derived their values from this shared object, eliminating independent parameter sources.
Immediately before the write, the workflow performed a fresh get_task_by_id call and compared the current title against the expected baseline stored in the envelope. Only a match allowed execution to continue. After the update, another read verified that the requested title now existed in Bitrix24. Acknowledgement from the API and the verified post-write state were stored as separate evidence objects.
The series of controlled runs confirmed that the bound path succeeded end-to-end while the stale-state and extractor-error conditions were handled safely. The researcher concluded that boolean approval flags alone cannot guarantee that the parameters shown to a human are the same parameters later sent to the target system.
Related articles
TenetBench Benchmark Tests LLMs for Compliance with Professional Ethical Codes
Researchers from the Institute of Information Security Issues at ISP RAS and the Institute of Cybersecurity and Information Protection at RANEPA have developed TenetBench, a benchmark designed to evaluate how well large language models adhere to specific norms in professional codes of ethics. The benchmark generates realistic work-related prompts that may conflict with rules from documents such as the Code of Professional Ethics for Advocates or journalistic standards on undisclosed advertising. Each test case pairs a conflicting request with a neutral counterpart to measure whether models refuse, warn, or still provide prohibited content. Evaluation separates task completion, norm violation detection, and presence of explicit refusals using an LLM-as-a-Judge approach validated against human annotators. The full dataset contains 3,433 prompt pairs derived from 15 professional codes, with a public subset of 2,235 pairs from 10 documents. Human validation confirmed 99.5% of generated examples as fully plausible work scenarios.
Solar to Showcase Solar AI Firewall for Business Protection Against AI Risks at SOC Forum 2026
GK Solar will present its new Solar AI Firewall solution at SOC Forum 2026 to help organizations control interactions with AI services and prevent confidential data leaks through employee prompts. The product addresses growing risks from staff sending corporate data to neural networks and AI agents gaining access to internal systems. The forum takes place on 27-28 October in Moscow's Lomonosov cluster, with the official launch of Solar AI Firewall scheduled during the event. Solar will also highlight practical risks of AI adoption, including data transmission to external services, use of public models by employees, and new attack surfaces in corporate infrastructure. The solution will be demonstrated as part of an integrated ecosystem for access, network, and data protection. In addition, the company plans to introduce JSOC ONE, a new line of Solar JSOC services tailored for mid-sized businesses that require 24/7 monitoring without building their own SOC.
Researcher Runs 100 Abliterated AI Agents Against His Own Accounts, Achieves Five Successful Breaches
A security researcher deployed roughly 100 autonomous AI agents based on abliterated open-source models to attempt unauthorized access to his online accounts over a five-hour period. The agents successfully compromised three accounts by exploiting vulnerabilities in legacy personal projects and two additional accounts through password reuse and credential stuffing. They also conducted 16 social engineering attempts, harvested and cross-verified personal data from public sources, and generated phishing pages targeting the researcher's Substack and Hacker News presence. Despite these successes, the agents failed to discover any zero-day vulnerabilities in third-party software or gain access to Tier-0 accounts such as Gmail, 1Password, or banking services. The experiment relied on abliterated versions of GLM-5.3 and DeepSeek V4 Flash models running on cloud GPUs, orchestrated via Codex CLI inside Docker containers with access to Railway infrastructure. Total compute cost reached approximately $210, equating to roughly $40 per successful account compromise. The researcher concluded that inexpensive swarms of such agents could soon enable low-skill attackers to systematically map and exploit an individual's entire online footprint.
How Censorship in Closed AI Models Reshapes Science: Hugging Face Releases Open Carbon-X DNA Model
Hugging Face CEO Clement Delangue announced the release of Carbon-X, an open 14.8 billion parameter language model designed for reading, analyzing, and generating DNA sequences. The announcement highlighted how shallow guardrails in closed commercial models blocked even proofreading a tweet discussing the topic. Commenters and researchers pointed out that these input/output filters lack context awareness, blocking legitimate scientific queries while motivated adversaries bypass them through rephrasing. Carbon-X was built on Qwen3-14B with a hybrid BPE and 6-mer tokenizer, domain mid-training on 302 billion tokens, and safety achieved through data curation rather than runtime censorship. The model achieves performance comparable to closed frontier systems at roughly 1/1800th the inference cost for large-scale genomic tasks. The case underscores growing asymmetry between closed corporate AI labs building wet labs and independent researchers restricted by superficial safety layers.