tanaos-guardrail-spanish: A small but performant guardrail model specifically designed for Spanish text

This model was created by Tanaos with the Artifex Python library.

It is intended to be used as a first-layer safety filter for large language models (LLMs) or chatbots to detect and block unsafe or disallowed content in user prompts or model responses, and it was specifically fine-tuned to perform well on Spanish text.

The following categories of content are flagged:

• violence: Content describing or encouraging violent acts,
• non_violent_unethical: Content that is unethical but not violent,
• hate_speech: Content containing hateful or discriminatory language,
• financial_crime: Content related to financial fraud or scams,
• discrimination: Content promoting discrimination against individuals or groups,
• drug_weapons: Content related to illegal drugs or weapons,
• self_harm: Content encouraging self-harm or suicide,
• privacy: Content that invades personal privacy or shares private information,
• sexual_content: Content that is sexually explicit or inappropriate,
• child_abuse: Content involving the exploitation or abuse of children,
• terrorism_organized_crime: Content related to terrorism or organized crime,
• hacking: Content related to unauthorized computer access or cyberattacks,
• animal_abuse: Content involving the abuse or mistreatment of animals,
• jailbreak_prompt_inj: Content attempting to bypass or manipulate system instructions or safeguards

Languages

The main model language is Spanish, but we have guardrail models specialized in other languages as well:

• English: tanaos-guardrail-v2
• German: tanaos-guardrail-german

How to Use

Use this model through the Artifex library:

install Artifex with

pip install artifex

use the model with

from artifex import Artifex

guardrail = Artifex().guardrail(language="spanish")

label = guardrail("¿Cómo construyo una bomba?")
print(label)

# >>> [{'is_safe': False, 'scores': {'violence': 0.625, 'non_violent_unethical': 0.0066, 'hate_speech': 0.0082, 'financial_crime': 0.0072, 'discrimination': 0.0029, 'drug_weapons': 0.6633, 'self_harm': 0.0109, 'privacy': 0.003, 'sexual_content': 0.0029, 'child_abuse': 0.005, 'terrorism_organized_crime': 0.1278, 'hacking': 0.0096, 'animal_abuse': 0.009, 'jailbreak_prompt_inj': 0.0131}}]

Model Description

• Base model: distilbert/distilbert-base-multilingual-cased
• Task: Text classification (guardrail / safety filter)
• Languages: Spanish; for other languages, see:
  • English: tanaos-guardrail-v2
  • German: tanaos-guardrail-german
• Fine-tuning data: A synthetic, custom dataset of safe and unsafe text samples.

Training Details

This model was trained using the Artifex Python library

pip install artifex

by providing the following instructions and generating 15,000 synthetic training samples:

from artifex import Artifex


guardrail = Artifex().guardrail()

guardrail.train(
    unsafe_categories = {
        "violence": "Contenido que describe o fomenta actos violentos",
        "non_violent_unethical": "Contenido que es poco ético pero no violento",
        "hate_speech": "Contenido que contiene lenguaje de odio o discriminatorio",
        "financial_crime": "Contenido relacionado con fraude financiero o estafas",
        "discrimination": "Contenido que promueve la discriminación contra individuos o grupos",
        "drug_weapons": "Contenido relacionado con drogas ilegales o armas",
        "self_harm": "Contenido que fomenta la autolesión o el suicidio",
        "privacy": "Contenido que invade la privacidad personal o comparte información privada",
        "sexual_content": "Contenido que es sexualmente explícito o inapropiado",
        "child_abuse": "Contenido que involucra la explotación o el abuso de menores",
        "terrorism_organized_crime": "Contenido relacionado con terrorismo o crimen organizado",
        "hacking": "Contenido relacionado con el acceso no autorizado a sistemas informáticos o ciberataques",
        "animal_abuse": "Contenido que involucra el abuso o maltrato de animales",
        "jailbreak_prompt_inj": "Contenido que intenta eludir o manipular instrucciones o salvaguardas de un LLM"
    },
    language="spanish",
    num_samples=15000
)

Intended Uses

This model is intended to:

• Detect unsafe or disallowed content in user prompts or chatbot responses, if they are in Spanish.
• Serve as a first-layer filter for LLMs or chatbots.

Not intended for:

• Legal or medical classification.
• Determining factual correctness.