RogueGPT

RogueGPT: A Controlled Stimulus Generation Framework for News Authenticity Research

[JOSS] [PyPI] [CI] [arXiv] [DOI] [License]

[RogueGPT - Controlled AI stimulus generation pipeline for fake news research]

If RogueGPT is useful for your research, please cite the JOSS paper and star the repository. GitHub's "Cite this repository" button in the sidebar gives the reference in APA and BibTeX.

Install: pip install roguegpt, or pip install "roguegpt[mcp]" with the MCP server. See Installation.

What This Is

RogueGPT is research instrumentation. It generates controlled, fully parameterised news fragments so that researchers can measure how people and detectors respond to synthetic news under experimental conditions.

It is the upstream stage of a three-part pipeline: CRED-1 identifies unreliable sources, RogueGPT produces the stimuli, and JudgeGPT collects human judgements that link back to the exact generation parameters.

Scope and safeguards. The framework produces short fragments for delivery inside studies. It has no publishing, posting, distribution, scheduling, or audience-targeting capability, and none is planned. Researchers supply their own API credentials, so generation stays attributable and the upstream providers' usage policies continue to apply. Human data collection through JudgeGPT is subject to institutional ethics review, the corpus is archived under restricted academic access, and the software is GPL-3.0 so derivative work stays open to inspection. See Limitations and Disclaimer.

Motivation

Empirical research on AI-generated news increasingly needs stimuli that are diverse, reproducible, and generated under controlled experimental conditions - not only for human-perception studies, but for the broader shift from static fake-news generation to agentic disinformation campaigns. RogueGPT provides a systematic, multi-model framework for producing such stimuli across multiple LLM families, languages, journalistic styles, and content formats, with full provenance tracking.

Unlike single-model predecessors (e.g., GROVER), RogueGPT uses an extensible, model-agnostic architecture that connects static text generation with dynamic agent-based simulation (ABS) of misinformation diffusion. By standardizing the "seeding" phase of misinformation simulations, it supports red-teaming of detection systems against diverse, parameter-rich content.

Intended users. The infrastructure targets four overlapping communities: (i) misinformation researchers who need controlled stimuli linked to perception data; (ii) red-teamers who generate adversarial content variants to stress-test detectors; (iii) agent-based simulation researchers who need reproducible seed content for diffusion models; and (iv) human-factors researchers studying perception.

RogueGPT is the stimulus generation component of the JudgeGPT research pipeline. Together, they enable end-to-end experiments from controlled content generation to quantitative human perception measurement. This methodology is described in our foundational survey, "Blessing or Curse? A Survey on the Impact of Generative AI on Fake News" (arXiv:2404.03021), and extended in two WWW '26 Companion papers: "Industrialized Deception" (arXiv:2601.21963), which examines the systemic effects of LLM-generated misinformation, and "Eroding the Truth-Default" (arXiv:2601.22871), which reports the human perception findings from stimuli generated by this framework.

Architecture

RogueGPT follows a three-layer architecture that separates data logic from interfaces:

┌──────────────────────────────────────────────────────┐│                   Interfaces                         ││  ┌───────────┐  ┌───────────┐  ┌──────────────────┐  ││  │  app.py   │  │  cli.py   │  │  mcp_server.py   │  ││  │ Streamlit │  │ Terminal  │  │  MCP Protocol    │  ││  └─────┬─────┘  └─────┬─────┘  └────────┬─────────┘  ││        └──────────────┼─────────────────┘            ││                  ┌────┴────┐                         ││                  │ core.py │                         ││                  │  Data   │                         ││                  │  Layer  │                         ││                  └────┬────┘                         ││               ┌───────┴────────┐                     ││               │    MongoDB     │                     ││               │ realorfake.    │                     ││               │ fragments      │                     ││               └────────────────┘                     │└──────────────────────────────────────────────────────┘

Dataset

The current corpus contains 3,278 multilingual news fragments spanning:

• 10 models across 6 providers, with per-model fragment counts: OpenAI (GPT-3.5 Turbo 480, GPT-4o 465, GPT-4.1 67, GPT-4 2), Google (Gemma 7B 454, Gemini 3 Flash 32), Meta (LLaMA-2 13B Chat 452), Anthropic (Claude Opus 4.6 282), Mistral (Mistral 7B Instruct v0.2 273), Microsoft (Phi-3 Mini 4k Instruct 131)
• 4 languages (English, German, French, Spanish)
• 3 formats (tweet, headline, short article)
• 5 journalistic styles per language (e.g., NYT, BBC, CNN, Fox News, WSJ for English)
• 640 human-sourced fragments as experimental anchors (legitimate news + fake-news references)

Of the full corpus, 2,638 fragments are machine-generated and 640 are human-authored, all with binary human/machine labels and full generation metadata (source model, prompt type, language, format, topic).

The corpus is available on Zenodo under restricted access for academic research. The machine-generated fragments are released under CC BY 4.0; the human-sourced fragments are excerpts of third-party news material that we cannot license onward and are shared for research use only, which is why access is granted on request rather than openly:

[DOI]

Citing a version

The concept DOI 10.5281/zenodo.18703137 always resolves to the latest version. The current version is 1.2.1, 10.5281/zenodo.23053786 (3,278 fragments, snapshot of 23 March 2026). Cite the version DOI when your analysis has to be reproducible.

Note for machine learning use: the corpus contains duplicate content under distinct fragment identifiers (545 rows across 196 repeated texts). Group by normalized text before creating train and test splits, otherwise a random split leaks training content into the test set.

The figures above describe the released corpus. The model registry is a separate, larger surface: prompt_engine.json defines 47 model identifiers across 11 providers as of 2026-02-23, enabling rapid expansion of the corpus with new model generations. Registry entries are available for generation; they do not imply that a model contributed to the current snapshot.

Both figures are reproducible from the repository:

# corpus: models and providers actually representedpython3 -c "import csv,collections; r=list(csv.DictReader(open('roguegpt-fragments.csv'))); m=[x['MachineModel'] for x in r if x['MachineModel'].strip()]; print(len(r),'fragments |',len(m),'machine |',len(set(m)),'models |',len({x.split('_')[0] for x in m}),'providers')"
# registry: models and providers configuredpython3 -c "import json; m=json.load(open('prompt_engine.json'))['GeneratorModel']; print(len(m),'models |',len({x.split('_')[0] for x in m}),'providers')"

Data Sourcing: Human-Written Fragments

Human-written fragments are sourced from two categories:

Legitimate News is collected from established, internationally recognized outlets via RSS feeds, including BBC News, The Guardian, Reuters, Tagesschau, FAZ, Le Monde, and El País. For each article, the opening paragraphs are extracted and stored alongside the source URL and outlet name for full provenance.

Fake News is identified using the CRED-1 Domain Credibility Dataset (Loth, Kappes & Pahl, 2025; DOI: 10.5281/zenodo.18769460), which scores 2,673 domains on a 0–1 credibility scale based on editorial classification, fact-check claims, web popularity, and domain age. Domains categorized as fake, unreliable, or conspiracy with a CRED-1 score below 0.25 and an active Tranco web rank serve as the source pool. Articles are then scraped from these domains' RSS feeds using the same extraction pipeline as legitimate news. This approach ensures that fake news fragments reflect authentic disinformation language rather than artificially constructed examples.

Both categories are ingested via the RogueGPT CLI with full metadata (--origin Human, --outlet, --url, and --is-fake for fake news), enabling reproducible dataset expansion.

Research Pipeline

RogueGPT operates as the first stage of a two-part experimental workflow:

1. Stimulus generation (RogueGPT): Fragments are produced with explicit control over model, style, language, format, and seed phrase. The model identifier, the full prompt and, for batch generation, the sampling parameters are persisted alongside the content. Fragments created through the web interface record model and prompt but not sampling settings, since that path uses provider defaults.
2. Storage (MongoDB): Each fragment is stored with full provenance metadata, enabling reproducible filtering by any experimental variable.
3. Human evaluation (JudgeGPT): Participants assess fragments on continuous dual-axis scales (source attribution and authenticity), producing quantitative perception data linked to generation parameters.
4. Analysis: The combined dataset supports investigations into model-specific detectability, cross-linguistic perception differences, and the role of individual differences in judgment accuracy.

Comparison with Existing Solutions

RogueGPT (generation) and JudgeGPT (evaluation) are positioned against notable fake-news generation systems along dimensions that matter for seeding agent-based simulations and red-teaming detectors. The comparison is feature-scoped (design intent and implemented functionality), not a claim of superior generation quality or detection performance.

Unlike single-model predecessors, RogueGPT's multi-model backend (GPT-4, LLaMA, Mistral, DeepSeek) enables heterogeneous agent networks where different nodes operate with varying capabilities, and its provenance tracking supports reproducible tracing of judgments back to generation parameters.

Installation

Prerequisites

• Python 3.10+
• MongoDB instance (local or Atlas)

From PyPI

pip install roguegpt            # command-line interface: roguegptpip install "roguegpt[mcp]"     # plus the MCP server: roguegpt-mcp

From source (needed for the Streamlit interface)

git clone https://github.com/aloth/RogueGPT.gitcd RogueGPTpython3 -m venv venvsource venv/bin/activatepip install -r requirements.txt

Configuration

Set the MongoDB connection string as an environment variable:

export ROGUEGPT_MONGO_URI="mongodb+srv://user:[email protected]/?retryWrites=true&w=majority"

Alternatively, for the Streamlit interface, configure .streamlit/secrets.toml:

[mongo]connection = "mongodb+srv://user:[email protected]/..."

Usage

Web Interface (Streamlit)

streamlit run app.py

Provides two modes: Generator (automated, template-driven generation across parameter combinations) and Manual Data Entry (for human-sourced fragments).

Command-Line Interface

The CLI enables scripted dataset operations without a browser. Installed from PyPI, the command is roguegpt; in a source checkout, python3 cli.py does the same:

# Dataset statisticspython3 cli.py stats
# List all configured model identifierspython3 cli.py models
# Retrieve random fragmentspython3 cli.py retrieve --n 5 --origin Machine --model openai_gpt-4o_2024-08-06
# Ingest a machine-generated fragmentpython3 cli.py ingest \  --origin Machine \  --model "openai_gpt-4.1" \  --lang en \  --is-fake \  --prompt "Write a short article about '''Topic''' in English in the style of CNN." \  --content "The generated text content..."
# Ingest a human-sourced fragmentpython3 cli.py ingest \  --origin Human \  --outlet "BBC" \  --url "https://www.bbc.com/news/example" \  --lang en \  --content "The original article text..."

Validation: The CLI enforces schema constraints from prompt_engine.json. Model identifiers are validated against the configuration by default. Use --lenient to allow unregistered models with a warning.

MCP Server (AI Agent Integration)

RogueGPT exposes its data layer via the Model Context Protocol for integration with AI agents and LLM tool-use workflows:

roguegpt mcp                  # installed with pip install "roguegpt[mcp]" (or: roguegpt-mcp)python3 mcp_server.py         # from a source checkout

Claude Desktop, Cursor or any MCP client:

{  "mcpServers": {    "roguegpt": {      "command": "uvx",      "args": ["--from", "roguegpt[mcp]", "roguegpt", "mcp"],      "env": { "ROGUEGPT_MONGO_URI": "mongodb+srv://..." }    }  }}

Tools:

Resources:

The MCP interface enables automated corpus expansion: an AI agent can generate content with any LLM, then ingest the result with full metadata for subsequent human evaluation.

Fragment Schema

Each fragment in the database conforms to the following schema:

Extending the Model Configuration

To add a new model, append its identifier to the GeneratorModel array in prompt_engine.json:

"GeneratorModel": [    "openai_gpt-4.1",    "anthropic_claude-opus-4-6",    "your-provider_model-name",    ...]

The naming convention follows provider_model-variant (e.g., openai_gpt-4o_2024-08-06, meta_llama-3.3-70b, anthropic_claude-sonnet-4-5). All ingestion interfaces validate against this list.

Roadmap

• Multimodal stimuli: Extend generation to images and multimedia for deepfake perception research.
• Automated corpus expansion: Agent-driven generation pipelines using the MCP server to systematically cover new model releases.
• Provenance integration: Content authenticity metadata (C2PA) annotation for mitigation experiments.
• Cross-dataset linking: Direct integration with JudgeGPT perception data for unified analysis.

Limitations

RogueGPT describes implemented functionality rather than a completed empirical evaluation; several limitations follow:

• Commercial API dependence: proprietary models (e.g., GPT-4) incur cost and tie reproducibility to third-party endpoints; local models (LLaMA, Mistral, DeepSeek) mitigate but do not remove this.
• Reproducibility across evolving models: providers update or deprecate versions, so identical prompts can drift over time; provenance metadata records the exact model/configuration used, but bit-exact regeneration is not guaranteed.
• Multilingual limits: cross-lingual quality depends on the underlying model and degrades for lower-resource languages.
• Human data collection: JudgeGPT requires participant recruitment and ethics/IRB approval; survey design needs validation for data quality.
• Misuse potential: the generation component can produce misinformation-like content; see the Disclaimer and intended research-only use.

For Researchers

Expert Survey

We are conducting a follow-up study to gather expert perspectives on AI-driven disinformation risks and mitigation strategies. If you have expertise in AI, policy, or journalism, we invite your participation:

Expert Survey (15 min)

All responses are treated confidentially and reported in anonymized, aggregated form.

Citation

If you use RogueGPT or its dataset in your work, please cite:

@article{loth2026roguegptjoss,    author    = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},    title     = {{RogueGPT}: A Controlled Stimulus Generation Framework for                 News Authenticity Research},    journal   = {Journal of Open Source Software},    year      = {2026},    volume    = {11},    number    = {125},    pages     = {11219},    publisher = {The Open Journal},    doi       = {10.21105/joss.11219},    url       = {https://doi.org/10.21105/joss.11219}}
@inproceedings{loth2026collateraleffects,    author    = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},    title     = {Industrialized Deception: The Collateral Effects of                 {LLM}-Generated Misinformation on Digital Ecosystems},    booktitle = {Companion Proceedings of the ACM Web Conference 2026                 (WWW '26 Companion)},    year      = {2026},    month     = jun,    publisher = {ACM},    address   = {New York, NY, USA},    location  = {Dubai, United Arab Emirates},    pages     = {295--302},    doi       = {10.1145/3774905.3795471},    url       = {https://arxiv.org/abs/2601.21963}}
@inproceedings{loth2026eroding,    author    = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},    title     = {Eroding the Truth-Default: {A} Causal Analysis of Human                 Susceptibility to Foundation Model Hallucinations and                 Disinformation in the Wild},    booktitle = {Companion Proceedings of the ACM Web Conference 2026                 (WWW '26 Companion)},    year      = {2026},    month     = jun,    publisher = {ACM},    address   = {New York, NY, USA},    location  = {Dubai, United Arab Emirates},    pages     = {1125--1132},    doi       = {10.1145/3774905.3795832},    url       = {https://arxiv.org/abs/2601.22871}}
@misc{loth2024blessing,    author        = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},    title         = {Blessing or Curse? {A} Survey on the Impact of Generative {AI}                     on Fake News},    year          = {2024},    doi           = {10.48550/arXiv.2404.03021},    eprint        = {2404.03021},    archivePrefix = {arXiv},    primaryClass  = {cs.CL}}
@inproceedings{loth2026killchaingi,    author        = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},    title         = {Interrupting the Chain: Human Perception of {AI}-Generated                     Disinformation Through a Kill Chain Lens},    booktitle     = {INFORMATIK 2026, Lecture Notes in Informatics (LNI)},    year          = {2026},    publisher     = {Gesellschaft f{\"u}r Informatik},    address       = {Bonn, Germany},    eprint        = {2608.21389},    archivePrefix = {arXiv},    primaryClass  = {cs.CY}}
@dataset{loth2026roguecorpus,    author    = {Loth, Alexander and Kappes, Martin and Pahl, Marc-Oliver},    title     = {RogueGPT Stimulus Corpus: A Multilingual                 {LLM}-Generated News Dataset},    year      = {2026},    publisher = {Zenodo},    doi       = {10.5281/zenodo.18703137},    url       = {https://doi.org/10.5281/zenodo.18703137}}

Related Publications

RogueGPT is part of a broader research program on generation and human perception of AI-generated news:

• Can Humans Tell? A Dual-Axis Study of Human Perception of LLM-Generated News (WebSci Companion '26). Reports human perception findings from RogueGPT-generated stimuli. doi:10.1145/3795513.3807431
• The Indistinguishability Threshold: Measuring Cognitive Vulnerabilities to AI-Generated Disinformation (WebSci Companion '26, PhD Symposium). doi:10.1145/3795513.3807421
• Industrialized Deception: The Collateral Effects of LLM-Generated Misinformation on Digital Ecosystems (WWW '26 Companion). doi:10.1145/3774905.3795471
• Eroding the Truth-Default: A Causal Analysis of Human Susceptibility to Foundation Model Hallucinations and Disinformation in the Wild (WWW '26 Companion). Uses this corpus as the stimulus set. doi:10.1145/3774905.3795832
• Interrupting the Chain: Human Perception of AI-Generated Disinformation Through a Kill Chain Lens (INFORMATIK 2026, LNI). arXiv:2608.21389
• Blessing or Curse? A Survey on the Impact of Generative AI on Fake News (2024). The survey this research program builds on. arXiv:2404.03021

Related Projects

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/new-model-support)
3. Commit your changes (git commit -m 'Add support for model X')
4. Push to the branch (git push origin feature/new-model-support)
5. Open a Pull Request

For substantial changes, please open an issue first.

License

This project is licensed under the GNU General Public License v3.0. See LICENSE for details.

Acknowledgments

This research is supported by Frankfurt University of Applied Sciences and IMT Atlantique. We thank the open-source communities behind Streamlit, MongoDB, and the Model Context Protocol for the infrastructure that makes this work possible.

Disclaimer

RogueGPT is an independent research project. The use of "GPT" in the project name follows pars pro toto convention, referring to the broader class of generative pre-trained transformer models. This project is not affiliated with or endorsed by OpenAI. All research adheres to established ethical guidelines for AI safety research.