# Responsible AI Labs

> Responsible AI Labs builds the RAIL Score — an 8-dimension responsible AI evaluation framework for scoring LLM outputs across Fairness, Safety, Reliability, Transparency, Privacy, Accountability, Inclusivity, and User Impact. Available as an API, Python SDK, JavaScript SDK, and web tools.

## Main Pages

- [Home](https://responsibleailabs.ai/): Platform overview, live RAIL Score demo, and product highlights.
- [About](https://responsibleailabs.ai/about): Mission, team, and background on the RAIL framework.
- [Pricing](https://responsibleailabs.ai/pricing): Free, Pro, Business, and Enterprise plans with credit-based usage.
- [Get Started](https://responsibleailabs.ai/get-started): Quick onboarding flow for new users.
- [RAIA](https://responsibleailabs.ai/raia): Responsible AI Assessment — enterprise evaluation service.
- [Contact](https://responsibleailabs.ai/contact-us): Contact form and support information.

## Tools

- [RAIL Evaluator](https://responsibleailabs.ai/tools/evaluator): Score any AI prompt/response pair across all 8 RAIL dimensions. Supports basic and deep evaluation modes.
- [Compliance Tester](https://responsibleailabs.ai/tools/compliance-tester): Check AI content against GDPR, CCPA, HIPAA, EU AI Act, India DPDP, and India AI Governance frameworks.
- [Protected Content](https://responsibleailabs.ai/tools/protected-content): Safe-regenerate AI responses that fail RAIL thresholds.

## Developer Documentation

Full API reference, SDK guides, and integration tutorials at [docs.responsibleailabs.ai](https://docs.responsibleailabs.ai).

- [Quickstart](https://docs.responsibleailabs.ai/getting-started/quickstart): Get your first RAIL Score in under 5 minutes.
- [Authentication](https://docs.responsibleailabs.ai/getting-started/authentication): API key setup and bearer token auth.
- [RAIL Framework](https://docs.responsibleailabs.ai/concepts/rail-framework): The 8 dimensions explained with scoring anchors (0–10).
- [Python SDK](https://docs.responsibleailabs.ai/sdk/python/overview): `pip install rail-score-sdk` — evaluate, compliance check, safe-regenerate, and agent evaluation.
- [JavaScript SDK](https://docs.responsibleailabs.ai/sdk/javascript/overview): `npm install @responsible-ai-labs/rail-score` — same capabilities for Node.js and browser.
- [API Reference — Evaluation](https://docs.responsibleailabs.ai/api-reference/evaluation): `POST /railscore/v1/eval` with basic and deep modes.
- [API Reference — Compliance](https://docs.responsibleailabs.ai/api-reference/compliance): `POST /railscore/v1/compliance/check` against 6 regulatory frameworks.
- [API Reference — Safe Regeneration](https://docs.responsibleailabs.ai/api-reference/safe-regeneration): `POST /railscore/v1/safe-regenerate`.
- [Agent Evaluation](https://docs.responsibleailabs.ai/sdk/python/agent-evaluation): Evaluate agent tool calls and results before and after execution.

## Knowledge Hub

Articles and guides at [knowledge.responsibleailabs.ai](https://knowledge.responsibleailabs.ai) and [responsibleailabs.ai/knowledge-hub](https://responsibleailabs.ai/knowledge-hub).

- [What is RAIL Score?](https://responsibleailabs.ai/knowledge-hub/articles/what-is-rail-score): Introduction to the 8-dimension responsible AI evaluation framework.
- [8 Dimensions of Responsible AI](https://responsibleailabs.ai/knowledge-hub/articles/8-dimensions-responsible-ai): Deep dive into Fairness, Safety, Reliability, Transparency, Privacy, Accountability, Inclusivity, User Impact.
- [Integrating RAIL Score with Python](https://responsibleailabs.ai/knowledge-hub/articles/integrating-rail-score-python): Step-by-step integration guide with LangChain, OpenAI, and CI/CD.
- [Building an Ethics-Aware Chatbot](https://responsibleailabs.ai/knowledge-hub/articles/building-ethics-aware-chatbot): FastAPI implementation with 5 threshold profiles.
- [EU AI Act Compliance 2025](https://responsibleailabs.ai/knowledge-hub/articles/eu-ai-act-compliance-2025): GPAI rules, timelines, and RAIL dimension mapping.
- [Financial Services AI Compliance](https://responsibleailabs.ai/knowledge-hub/articles/financial-services-ai-compliance): SR11-7/ECOA mapping and multi-layer compliance stack.
- [LLM Evaluation Benchmarks 2025](https://responsibleailabs.ai/knowledge-hub/articles/llm-evaluation-benchmarks-2025): Comparison of evaluation frameworks and methodologies.
- [Engineering Alignment with RAIL-HH-10K](https://responsibleailabs.ai/knowledge-hub/articles/engineering-alignment-rail-hh-10k): Dataset design, DeBERTa fine-tuning, and alignment results.

## Open Source

Public SDKs, integrations, datasets, and research at [responsibleailabs.ai/open-source](https://responsibleailabs.ai/open-source).

- [Python SDK (rail-score-sdk)](https://pypi.org/project/rail-score-sdk/): Official Python client on PyPI (MIT). `pip install rail-score-sdk`. First release 2025-10-18. Source: [github.com/Responsible-AI-Labs/rail-score-sdk](https://github.com/Responsible-AI-Labs/rail-score-sdk).
- [JavaScript/TypeScript SDK (@responsible-ai-labs/rail-score)](https://www.npmjs.com/package/@responsible-ai-labs/rail-score): Official JS/TS client on npm (MIT). `npm install @responsible-ai-labs/rail-score`. First release 2025-11-03. Source: [github.com/Responsible-AI-Labs/rail-score-js](https://github.com/Responsible-AI-Labs/rail-score-js).
- [Drupal module (rail_score)](https://www.drupal.org/project/rail_score): Drupal 9/10/11 integration. `composer require drupal/rail_score`. Source: [github.com/Responsible-AI-Labs/rail-score-drupal](https://github.com/Responsible-AI-Labs/rail-score-drupal).
- [RAIL paper on arXiv](https://arxiv.org/abs/2505.00204): Peer-reviewed methodology paper.
- [HuggingFace org](https://huggingface.co/responsible-ai-labs): Public datasets including RAIL-HH-10K.
- [GitHub organization](https://github.com/Responsible-AI-Labs): All public repositories.

## AI Incident Watch

Public, fact-checked catalogue of real-world AI harm incidents from 2021 onwards at [responsibleailabs.ai/ai-watch](https://responsibleailabs.ai/ai-watch). Every entry is cross-checked against at least two independent top-tier sources (major news outlets, regulators, court filings, or NGO investigations).

- 39 verified incidents covering 14 categories (chatbot-linked deaths, facial-recognition wrongful arrests, deepfake fraud, autonomous-vehicle harm, algorithmic discrimination, AI hallucination, AI voice cloning scams, deepfake political disinformation, deepfake image-based sexual abuse, data privacy violations).
- Coverage by region: India (Rashmika Mandanna deepfake, Sachin Tendulkar Skyward deepfake, Ratan Tata investment-scam ads, 2024 Lok Sabha AI election wave, Kerala WhatsApp deepfake fraud), United States (Character.AI and ChatGPT and Gemini wrongful-death suits including Sewell Setzer III, Adam Raine, Zane Shamblin, Jonathan Gavalas; Detroit and Fargo facial-recognition wrongful arrests; NYC MyCity chatbot; Tesla Autopilot fatalities; Cruise robotaxi pedestrian drag), Europe (Dutch Toeslagenaffaire, Clearview AI EU fines, Slovakia 2023 election deepfake, Italy Replika ban, Almendralejo nudify case), Asia-Pacific (South Korea Telegram deepfake crisis, Indonesia Suharto deepfake, Australian Robodebt, Hong Kong Arup CFO deepfake scam), and Other regions (Israel Red Wolf surveillance, Las Vegas cartel virtual-kidnapping voice clone, Air Canada chatbot hallucination).
- Each detail page lists a verified summary, impact, outcome, location, AI system, responsible entities, victims, lives lost, financial loss, and source URLs with publisher names.
- Listing default sort pins fatal incidents and Indian cases to the top, then date desc.
- Page features a world map (choropleth fill by incident count, more red = more harm), a region tab bar (All / India / United States / Europe / Asia-Pacific / Other), category / year / harm-scope filters, free-text search, and grid/list view toggle.
- Crisis hotlines (India iCall 9152987821 and Vandrevala Foundation listed first, then US 988, UK Samaritans, Australia Lifeline, Canada 988, Brazil CVV, global findahelpline.com directory), regulator complaint routes by region (MeitY cybercrime, India NCW, RBI Sachet, US FTC reportfraud, US EEOC, FBI IC3, EU DPAs, noyb), and advocacy organisations (Internet Freedom Foundation, ACLU, EFF, Algorithmic Justice League, Privacy International, Lighthouse Reports, Amnesty International, AI Now Institute) are linked from the page.

## Announcements

Press releases, partnerships, and product updates at [responsibleailabs.ai/announcements](https://responsibleailabs.ai/announcements). Also includes a downloadable press kit (logos, contact).

- [RAIL joins Google for Startups Cloud, AWS, MongoDB, and Atlassian startup programs](https://responsibleailabs.ai/announcements/startup-programs-2026) (April 22, 2026): Infrastructure and tooling partnerships behind the RAIL Score platform.
- [Responsible AI Labs selected for Nasscom GenAI Foundry Cohort 4](https://responsibleailabs.ai/announcements/nasscom-genai-foundry-cohort-4) (April 17, 2026): Selected as one of 33 high-potential Indian GenAI startups inducted into Nasscom's fourth GenAI Foundry cohort. Covered by The Hindu.
- [RAIL datasets on HuggingFace](https://responsibleailabs.ai/announcements/huggingface-datasets-2026) (March 12, 2026): RAIL-HH-10K (multi-dimensional safety dataset, 789 downloads) and the Indian Responsible AI Benchmark (158 downloads).
- [Updated API Endpoints & SDK v2.1](https://responsibleailabs.ai/announcements/api-sdk-v2-1) (March 7, 2026): Compliance testing endpoints, safe content regeneration, and expanded Python/JavaScript SDK support.
- [First Stable Release, RAIL Score API v1.0.0](https://responsibleailabs.ai/announcements/rail-score-v1-0-0) (November 15, 2025): Unified eval and safe-regenerate endpoints at api.responsibleailabs.ai.
- [RAIL Score goes open source: Python, JavaScript/TypeScript, and Drupal](https://responsibleailabs.ai/announcements/rail-sdks-public-2025) (November 4, 2025): All three official client libraries are now public and MIT-licensed.
- [Our research paper is on arXiv: RAIL in the Wild](https://responsibleailabs.ai/announcements/arxiv-rail-in-the-wild-2025) (April 30, 2025): arXiv:2505.00204 operationalizing the eight-dimension RAIL framework on Anthropic's Values in the Wild dataset (308K+ Claude conversations, 3K+ annotated value expressions).

## Legal

- [Privacy Policy](https://responsibleailabs.ai/privacy-policy)
- [Terms of Service](https://responsibleailabs.ai/terms-of-service)
- [Responsible AI Policy](https://responsibleailabs.ai/responsible-ai-policy)

## Optional: Full Content Index

For a complete article-by-article index, see [llms-full.txt](https://responsibleailabs.ai/llms-full.txt).